EPA CompTox Research Program FY2009 to 2012
Final
COMPUTATIONAL
1 U 1 0 0 1 1 0 0 1 I U 0 0 I U u
? i ?; TOXICOLOGY
U.S. EPA Office of Research and Development
Computational Toxicology Research Program
Implementation Plan for Fiscal Years 2009 to 2012
Providing High-Throughput Decision Support Tools for Screening and
Assessing Chemical Exposure, Hazard, and Risk
This document has been reviewed by the United States Environmental Protection Agency,
Office of Research and Development (ORD) and approved for public release, but it does not
necessarily constitute official Agency policy. This plan follows the first generation Computational
Toxicology Research Program Implementation Plan for Fiscal Years 2005 to 2008, in providing
a strategic overview of research for fiscal years 2009 to 2012. This plan was reviewed by ORD
senior management and members of the Science Council, as well as by the Computational
Toxicology Subcommittee of the ORD Board of Scientific Counselors on September 29 and 30,
2009, in Research Triangle Park, NC.
-------
EPA CompTox Research Program FY2009 to 2012
Final
Table of Contents
List of Figures iv
List of Acronyms v
Executive Summary vii
I. History of the CTRP and the NCCT 1
A. Defining the Mission of Computational Toxicology at EPA 1
B. Timeline of CTRP Development 2
C. Resources for the CTRP 4
D. BOSC Reviews of the CTRP 5
E. NRC Report and EPA's Strategic Plan for 21st Century Toxicology 6
1. NRC Report on Toxicity Testing in the 21st Century 7
2. EPA Strategic Plan for Evaluating the Toxicity of Chemicals 7
F. Significant Accomplishments of the CTRP: FY2006 to 2008 9
1. Accomplishments of the NCCT 9
2. Accomplishments by the CTRP ORD Intramural Partners 11
3. Accomplishments by STAR Grantees in the CTRP 14
G. Retrospective Summary of the CTRP and NCCT 15
II. Revision of the CTRP for FY2009 to 2012 16
A. Maturation of the Program 16
B. CTRP Integration Across ORD Laboratories and Centers 19
C. Regional and Program Interactions 23
D. Priority Areas for CTRP Management 25
1. Toxicity Predictions and Chemical Prioritizations Incorporating Exposure 25
2. Strengthening Cross-ORD Collaborations 26
3. Tox21: A Federal Partnership Transforming Toxicology 27
4. Communicating Computational Toxicology 28
a. EPA Program Office Training and Implementation of Computational Tools 28
b. Communities of Practice for Chemical Prioritization and Exposure Science 28
5. Developing Clients for Virtual Tissues 29
III. CTRP Project Summaries for FY2009 to 2012 30
A. Intramural Projects Coordinated by NCCT 30
1. ACToR—Aggregated Computational Toxicology Resource 30
2. DSSTox—Chemical Information Technologies in Support of Toxicology Modeling 31
3. ToxRefDB—Toxicity Reference Database 31
4. ChemModel— Application of Molecular Modeling to Assessing Chemical Toxicity 32
5. ToxCast™—Screening and Prioritization of Environmental Chemicals Based on
Bioactivity Profiling and Predictions of Toxicity 32
6. ExpoCast™—Exposure Science for Screening, Prioritization, and Toxicity Testing 33
7. v-Embryo™—Virtual Embryo 33
8. v-Liver™ —The Virtual Liver Project 34
9. Uncertainty— Analysis in Toxicological Modeling 34
B. Intramural Projects Coordinated by NERL and NHEERL 34
1. NERL 34
2. NHEERL 36
C. Extramural STAR Grantee Projects 37
D. Summary Integration of the CTRP Projects for FY2009 to 2012 38
IV. APPENDIXES 40
A. Intramural CTRP Projects 40
1. Project Plans 40
a. ACToR—Aggregated Computational Toxicology Resource 40
li
-------
EPA CompTox Research Program FY2009 to 2012 Final
b. DSSTox—Chemical Information Technologies in Support of Toxicology Modeling....42
c. ToxRefDB—Toxicity Reference Database 45
d. ChemModel—The Application of Molecular Modeling to Assessing Chemical Toxicity
47
e. ToxCast™— Screening and Prioritization of Environmental Chemicals Based on
Bioactivity Profiling and Predictions of Toxicity 49
f. ExpoCast™—Exposure Science for Screening, Prioritization, and Toxicity Testing...52
g. v-Embryo™—Virtual Embryo 55
h. v-Liver™ —The Virtual Liver Project 61
i. Uncertainty —Analysis in Toxicological Modeling 64
2. Project Outcomes - Providing High-Throughput Computational Tools for the Identification
of Chemical Exposure, Hazard, and Risk 69
B. Extramural STAR Centers Projects 73
C. FY2004 "New Start" Award Bibliography 75
iii
-------
EPA CompTox Research Program FY2009 to 2012
Final
List of Figures
Figure 1 2
Figure 2 4
Figure 3 5
Figure 4 20
Figure 5 22
Figure 6 30
iv
-------
EPA CompTox Research Program FY2009 to 2012
Final
List of Acronyms
ACToR
Aggregated Computational Toxicology Resource
BBDR
biologically based dose-response
BE
biomonitoring equivalent
BfR
Federal Institute for Risk Assessment
BOSC
Board of Scientific Counselors
CCL
Candidate Contaminant List
CDC
Centers for Disease Control and Prevention
CEBS
NIEHS Chemical Effects in Biological Systems
ChEBI
Chemical Entities of Biological Interest
CoP
communities of practice
CTISC
Computational Toxicology Implementation and Steering Committee
CTRP
Computational Toxicology Research Program
DNT
developmental neurotoxicity
DSST ox
Distributed Structure-Searchable Toxicity Database
ECOTOX
ECOTOXicology Database
EDC
endocrine-disrupting compound
EDSP
endocrine disruptor screening program
EMAGE
Edinburgh Mouse Atlas of Gene Expression
EPA
U.S. Environmental Protection Agency
ExpoCast™
Exposure Forecasting Project
ExpoCoP
Exposure Science Community of Practice
FDA
U.S. Food and Drug Administration
FTE
full-time equivalent
FTTW
Future of Toxicity Testing Workgroup
FY
fiscal year
HEASD
Human Exposure and Atmospheric Sciences Division
HEDS
Human Exposure Database System
HPV
High-Production Volume
HTS
high-throughput screening
ICCA-LRI
International Council of Chemical Association's - Long-Range Research
Initiative
IRIS
Integrated Risk Information System
KB
knowledge base
M eta Path
a metabolism pathways expert system
MICA
Mechanistic Indicators of Childhood Asthma
MOA
mode or mechanism of action
MOU
memorandum of understanding
MTA
material transfer agreement
MYP
multiyear plan
NCCT
National Center for Computational Toxicology
NCEA
National Center for Environmental Assessment
NCER
National Center for Environmental Research
NCGC
NIH Chemical Genomics Center
NERL
National Exposure Research Laboratory
NHEERL
National Health and Environmental Effects Research Laboratory
NHGRI
NIH Chemical Genome Research Institute
V
-------
EPA CompTox Research Program FY2009 to 2012
Final
NIEHS
National Institute of Environmental Health Sciences
NIH
National Institutes of Health
NLM
National Library of Medicine
NPD
National Program Director
NRC
National Research Council
NRMRL
National Risk Management Research Laboratory
NTP
National Toxicology Program
OECD
Organization for Economic Cooperation and Development
OPP
Office of Pesticide Programs
OPPT
Office of Pesticides, Prevention, and Toxics
OPPTS
Office of Pesticides, Prevention, and Toxic Substances
ORD
Office of Research and Development
OSCP
Office of Science Coordination and Policy
OW
Office of Water
PBPK
physiologically based pharmacokinetic
PMRA
Pest Management Regulatory Agency
QA
quality assurance
QMP
quality management plan
QSAR
quantitative structure-activity relationship
RIVM
The National Institute for Public Health and the Environment, Netherlands
RTD
NHEERL Reproductive Toxicology Division
SAB
Science Advisory Board
SAP
Scientific Advisory Panel
SAR
structure-activity relationship
SHEDS
Stochastic Human Exposure and Dose Simulation
STAR
Science To Achieve Results
TIVS
Texas-Indiana Virtual STAR Center
ToxCast™
Toxicity Forecasting Project
ToxML
Toxicity Databases Data Management Leadscope
ToxRefDB
Toxicity Reference Database
UMDNJ
University of Medicine and Dentistry of New Jersey
UNC
University of North Carolina
v-Embryo™
Virtual Embryo Project
v-Liver™
Virtual Liver Project
v-Liver-KB
v-Liver Knowledgebase
VT
virtual tissue
WHO
World Health Organization
vi
-------
EPA CompTox Research Program FY2009 to 2012
Final
Executive Summary
This document lays out the fiscal year 2009 to 2012 objectives of the U.S. Environmental
Protection Agency (EPA), Office of Research and Development (ORD) research program in
computational toxicology. Computational toxicology is the application of mathematical and
computer models to help assess chemical hazards and risks to human health and the
environment. Supported by advances in informatics, high-throughput screening technologies,
and systems biology, EPA is developing robust and flexible computational tools that can be
applied to the thousands of chemicals in commerce and the contaminant mixtures found in
America's air, water, and hazardous-waste sites. The ORD Computational Toxicology Research
Program (CTRP) is composed of three main elements. The largest component is the National
Center for Computational Toxicology (NCCT), which was established in 2005 to coordinate
research on chemical screening and prioritization, informatics, and systems modeling. The
second element consists of related activities in the National Health and Environmental Effects
Research Laboratory (NHEERL) and the National Exposure Research Laboratory (NERL). The
third and final component consists of academic centers working on various aspects of
computational toxicology and funded by the EPA Science to Achieve Results (STAR) program.
Together, these elements form the key components in the implementation of both the initial
strategy, A Framework for a Computational Toxicology Research Program (US EPA, 2003), and
the newly released The U.S. Environmental Protection Agency's Strategic Plan for Evaluating
the Toxicity of Chemicals (US EPA, 2009). Key intramural projects of the CTRP include
digitizing legacy toxicity testing information Toxicity Reference Database (ToxRefDB), predicting
toxicity (ToxCast™) and exposure (ExpoCast™), and creating virtual liver (v-Liver™) and virtual
embryo (v-Embryo™) systems models. EPA funded STAR centers also are providing
bioinformatics, computational toxicology data and models, and developmental toxicity data and
models. All CTRP projects participate in the Agency's formal quality assurance program and
regularly undergo peer review. The models and underlying data are being made available
publicly through the Aggregated Computational Toxicology Resource (ACToR), the Distributed
Structure-Searchable Toxicity (DSSTox) Database Network, and other EPA websites. Thus, the
CTRP is providing the foundation for advancing high-throughput toxicology and risk
assessments and, thereby, closing critical data gaps for thousands of chemicals and helping
EPA better assess and manage chemical risk.
The CTRP is evolving beyond the initial focus on hazard identification and chemical
prioritization, as expressed in the new long-term goal of providing high-throughput decision
support tools for assessing chemical exposure, hazard, and risk. There is an increasing
emphasis on using high-throughput bioactivity profiling data in systems modeling to support
quantitative risk assessments and greater involvement in developing complementary higher
throughput exposure models. Discussions are well underway among NCCT, NHEERL, NERL,
the National Risk Management Research Laboratory, the National Center for Environmental
Assessment, and the National Center for Environmental Research, managers of the STAR
program on how the CTRP will play a major role in future integrated ORD programs centered on
looking at chemical hazards and risks from a life cycle viewpoint. This integrated approach will
enable analysis of life stage susceptibility and understanding of the exposures, pathways, and
key events by which chemicals exert their toxicity in developing systems (e.g., endocrine related
pathways). The CTRP will be a critical component in next generation risk assessments utilizing
quantitative high-throughput data and providing a much higher capacity for assessing chemical
toxicity than is currently available.
This second generation CTRP implementation plan is highly consistent with the Agency's
priority for improving the management of chemical and contaminant risks. In January 2009, a
-------
EPA CompTox Research Program FY2009 to 2012
Final
memo sent to all EPA employees by Administrator Jackson listed managing chemical risks as
one of five top priorities and stated:
"More than 30 years after Congress enacted the Toxic Substances Control Act, it
is clear that we are not doing an adequate job of assessing and managing risks
of chemicals in consumer products, the workplace and the environment. It is now
time to revise and strengthen EPA's chemicals management and risk
assessment programs."
With contributions from across ORD, the CTRP will provide EPA program offices better decision
analysis tools for hazard and exposure screening and assessment, which then can be used to
better manage the risks of chemicals. The CTRP is acquiring an international reputation for
leadership in the introduction of innovative high-throughput technologies and computational
approaches for identifying toxicity pathways and characterizing response to environmental
exposures. It is through this effort that problems will be addressed and solutions to EPA's
chemicals management and risk assessment programs will be developed.
viii
-------
EPA CompTox Research Program FY2009 to 2012
Final
I. History of the CTRP and the NCCT
A. Defining the Mission of Computational Toxicology at EPA
Computational toxicology applies mathematical
and computer models and molecular biological
and chemical approaches to explore both
qualitative and quantitative relationships
between chemical exposure and adverse health
outcomes. Recent technological advances
make it possible to develop molecular profiles
using high-throughput and high-content
methods that identify the impacts of
environmental exposures on living organisms.
With these tools, scientists can produce a more detailed understanding of the hazards and risks
of a much larger number of chemicals. The integration of modern computing with molecular
biology and chemistry is enabling scientists to better understand a chemical's progression
through the environment to the target tissue within an organism and, ultimately, to the key steps
that trigger an adverse health effect. Currently, risk estimates most often are based on gross
outcomes of disease, such as occurrence of cancer, a neurological disorder, or a visible birth
defect. The National Research Council (NRC) in its 2007 report, Toxicity Testing in the 21st
Century: A Vision and a Strategy, called for a concerted effort to move toxicology from a
primarily descriptive science to a more predictive one by utilizing largely human based in vitro
studies to understand the biological pathways by which chemically induced diseases occur. The
Environmental Protection Agency's Computational Toxicology Research Plan is working to aid
this transformation by evaluating the key molecular changes occurring in the function of critical
human toxicity pathways within cells, tissues, individuals, and populations. The key will be
connecting these changes quantitatively and systemically to the types of adverse health effects
that have been the traditional basis of EPA risk assessments and to use this understanding to
reduce the current uncertainties in the extrapolation of effects across dose, species, and
chemicals. The Office of Research and Development CTRP is composed of three main
elements. The largest component is the National Center for Computational Toxicology, which
was established in 2005 to coordinate research on chemical screening and prioritization,
informatics, and systems modeling. The second element consists of related activities in the
National Health and Environmental Effects Research Laboratory (NHEERL) and National
Exposure Research Laboratory (NERL) of ORD. The third and final component consists of
academic centers working on various aspects of computational toxicology and funded by the
EPA Science to Achieve Results (STAR) Program.
The rapid and ongoing success of the CTRP is impacting hazard and exposure identification
and helping to close data gaps, identify toxicity pathways, suggest modes of action, and make
for more efficient utilization of precious resources on the highest priority chemicals. Besides
these initial outcomes from the higher throughput approach of the CTRP, informatics and
modeling efforts will provide more in-depth and quantitative molecular understanding of how
biological systems respond to environmental chemicals. These knowledgebases and in silico
tools will reduce or quantify uncertainties relating to biological susceptibility, species differences,
and dose response as part of a faster and more intelligent targeted testing paradigm in support
of quantitative risk assessments.
CTRP Mission Statement: To
integrate modern computing and
information technologies with
molecular biology to provide the
Agency with decision support tools
for high-throughput risk assessment.
1
-------
EPA CompTox Research Program FY2009 to 2012 Final
B. Timeline of CTRP Development
In fiscal year 2002, Congress ordered a redirection of $4 million from available EPA funds,
"...for the research, development and validation of non-animal alternative
chemical screening and prioritization methods, such as rapid, non-animal
screens and Quantitative Structure Activity Relationships (QSAR), for potential
inclusion in EPA's current and future relevant chemical evaluation programs."
To fulfill this directive, EPA embarked on development of a research program that (1) was
consistent with the Congressional mandate, (2) complemented and leveraged related ongoing
Agency sponsored efforts to consider alternative test methods, (3) further advanced the
research to support the Agency's mission, and (4) would not duplicate the mission and
programs in this area conducted by other agencies (see Figure 1 for a timeline of CTRP
development).
ORD Computational Toxicology
Research Program Development
Congressional Redirect
EDC Proof of Concepts
1st CTRP
CTRP Framework Implementation Plan
CTISC ToxCast Design
Seven CTRP Proof of Concepts 1st and 2nd STAR Centers
STAR Systems Biology RFA BOSC II
NCCT Staffing Complete
NAS 21st Century Vision
EPA 21st Century Strategy
Tox21 MOU
DSSTox v2
3rd STAR Center
BOSC III
FY02
FY04
FY06
FY08
FY03
FY05
FY07
FY09
CTRP Design Team
SAB and BOSC reviews
RTP CTRP Workshop
STAR HTS RFA
NCCT Launch
DSSTox v1
ToxCast Concept
BOSC I
1st Title 42 Hire
Chemical Prioritization
CoP
ToxCast Launch
v-Tissues Launch
Intl Science Forum
2nd CTRP
Implementation Plan
ACToR Launch
ExpoCast Launch
ToxRefDB Launch
ToxCast I Complete
1st ToxCast Summit
ToxCast II Launch
v-Tissues 09
4th STAR Center
BOSC IV
Figure 1
Thus the CTRP was initiated to target these goals and, in the process, significantly advance
toxicology and risk assessment as currently practiced by the Agency and the broader
environmental sciences community. In FY2002 to 2003 pilot projects were funded to
demonstrate computational toxicology could be adapted to the study of endocrine disruptors.
Early successes of these efforts included refinement of estrogen receptor ligand binding data for
development of quantitative structure-activity models, evaluation of EPA-developed cell lines for
2
-------
EPA CompTox Research Program FY2009 to 2012
Final
detecting estrogen and androgen activities from various species and the development of an
alternative test method for evaluating effects on steroidogenesis in H295R cells (see EDSP
Assay Status).
With increasing attention to and expectations for the CTRP in FY2003, ORD developed A
Framework for a Computational Toxicology Research Program, which was published in FY2004
and provided strategic direction for the program. This document was the product of a cross-
ORD design team of scientists and was endorsed by the Science Advisory Board (SAB). ORD
hosted a workshop in Research Triangle Park, NC, in late FY2003 to introduce the CTRP
framework (Kavlock et al., 2003) from which the three objectives for EPA computational
toxicology were translated into the three initial long-term goals (LTGs) for the program:
1) risk assessors use improved methods and tools to better understand and describe the
linkages of the source-to-outcome paradigm,
2) EPA program offices use advanced hazard characterization tools to prioritize and screen
chemicals for toxicological evaluation, and
3) EPA assessors and regulators use new and improved methods and models based on the
latest science for enhanced dose-response assessment and quantitative risk assessment.
With issuance of the CTRP framework, ORD began the process of implementing a more
formalized program. A cross-Agency working group, the Computational Toxicology
Implementation Steering Committee (CTISC) was formed in FY2004 to oversee the selection
and funding of projects across ORD. Seven cross-ORD projects were initiated as result of
CTISC action, and these "new start" projects became a critical component of the first generation
CTRP implementation plan. Greater detail on the accomplishments of these seven projects is
provided in Section I.F.2.
In October 2004, then EPA Science Advisor and Assistant Administrator for ORD, Dr. Paul
Gilman, announced the formation of the NCCT, which began official functions in February 2005.
The announcement states:
"The Center will advance the science needed to more quickly and efficiently
evaluate the potential risk of chemicals to human health and the environment.
The Center will coordinate and implement EPA's research on computational
toxicology to provide tools to conduct more rapid risk assessments and improve
the identification of chemicals for testing that may be of greatest risk."
NCCT quickly became the hub for ORD CTRP research. NCCT formed key partnerships with
the other laboratories and centers within ORD, which formed the second critical element.
Partnerships with NHEERL, NERL, the National Risk Management Research Laboratory, and
the National Center for Environmental Assessment helped in the execution of not only the seven
cross-ORD "new start" projects awarded by the CTISC in 2004, but also in several original
NCCT-led projects, including the Distributed Structure-Searchable Toxicity Database (DSSTox),
Toxicity Forecasting Project (ToxCast™),Toxicity Reference Database (ToxRefDB), and the
virtual tissues projects looking at liver and embryo. Greater detail on the accomplishments and
future plans for these and other NCCT projects is provided in Section I.F.1.
The third critical component of the CTRP is extramural partners and research, much of which is
supported by NCER through the STAR program. In FY2003 and 2004, two separate STAR
Requests for Applications (RFA) that funded projects in high-throughput screening and systems
biology were issued. In FY2006, two STAR academic centers to support the advancement of
bioinformatics in environmental health were funded, with a third center for computational
3
-------
EPA CompTox Research Program FY2009 to 2012
Final
toxicology funded in FY2008. An award for a fourth center, which will focus on pathways and
models of developmental toxicity, was made in late FY2009. Additional information on the STAR
centers, their accomplishments, and future plans is provided in Section I.F.3.
In ORD's Computational Toxicology Research Program Implementation Plan for FY2006-2008,
these various research efforts supporting the three LTGs identified in the CTRP framework were
grouped into five research tracks: (1) Development of Data for Advanced Biological Models; (2)
Information Technologies Development and Application; (3) Prioritization Method Development
and Application; (4) Providing Tools and System Models for Extrapolation Across Dose, Life
Stage, and Species; and (5) Advanced Computational Toxicology Approaches to Improve
Cumulative Risk Predictions. Within these five tracks, all of the NCCT, ORD intramural, and
STAR funded extramural project plans and interconnections were defined in this first-generation
implementation plan. The next generation of the CTRP implementation plan carries forward
many of these same research components into FY2009 to 2012.
As noted in the U.S. EPA Strategic Plan for Evaluating the Toxicity of Chemicals (see Section
I.E.2), many environmental statutes require that EPA consider both human and ecological
health risks, and although the initial emphasis of the CTRP has been on improvements in
human health assessments to provide the critical mass and resources necessary to be
successful, the program also, over time, must address ecological concerns as well. As we look
to the future, we see opportunities to leverage existing efforts such as the Tox21 project (see
Section II.D.3) and research on alternative species (e.g., the zebrafish projects underway at
NHEERL and at the newest STAR Center) to help transition the CTRP to a more balanced
human and ecological assessment program.
C. Resources for the CTRP
Funding of the CTRP has been relatively stable over the past several years, and this has
enabled the program to develop consistent with the strategic plan. In FY2009, the program was
funded at ~$15 million and 32 FTEs.
Approximately 50% of the resources are
allocated to NCCT, 25% to the STAR
program, and the remainder to NHEERL
and NERL. The majority of the FTEs
(-22) are located in NCCT. Figure 2
displays the history of the budget through
the President's FY2010 request, which
would provide an increase in funding,
initially to support Phase II of ToxCast™,
but then broader aspects of the program
in the out years.
40
35
30
254
20 4
15
10
5
0
MbME
S
~ FTEs
~ $ in Millions
j &
Figure 2
NCCT is organized into three primary functional groups: (1) Chemical Prioritization, (2) Systems
Modeling, and (3) Informatics, with a small group of administrative support personnel (Figure 3).
In addition to the permanent Federal staff of 23, there are a number of postdoctoral fellows,
predoctoral, student contractors, and Senior Environmental Enrollees a grantee Organization)
who support various aspects of the program. In 2006 and 2007 NCCT successfully recruited
three senior-level Title 42 scientists in bioinformatics and systems biology. These hires have
proven critical for NCCT to establish core research projects in predictive and systems toxicology
4
-------
EPA CompTox Research Program FY2009 to 2012 Final
and the necessary informatics and computational infrastructure. More recently, in 2008, two
research positions were filled with scientists coming through the NCCT postdoctoral program.
Full Curriculum Vitas of the staff are provided at the NCCT website. The NCCT currently is
recruiting a public affairs specialist to improve both internal and external communications.
ORD NCCT Organizational Chart
EPA Environmental Protection Agency
ORD Office of Research arid Development
NCCT
National Center for
Computational
Toxicology
Director
Robert Kaviock
Deputy Director (on detail)
Jerry Blancato
Administration
Management Analyst
Karen Dean
Program Analyst
Sandra Roberts
Dorothy Goodson
SEE Grantee
Public Affairs Specialist
TBD
Chemical Prioritization
David Dix (Acting Deputy Director]
Danie Rotroff - UNC Graduate Student
Keith Houck
Stephen Little
Matthew Martin
Inttiirany Thil ainadara.ah -
Technical SEE Grantee
James Rabinowitz
Systems Modeling
Elaine Cohen Hubal
Tom Knudsen
M cole K einsfruei e - Pas! Doc
Michael Rountree - Student Contractor
Anw Sirvqti - Locteed /Wart/71 Contactor
Nisha Snipes - Post Doc
R. Woodrow Setzer
Jimena L. Davis - Cross ORD Post Doc
Imran Shah
John Jack - Post Doc
John Wambauqh
J
Informatics
Richard Judson
Andrew Beam - NCSU Graduate Student
Hol.lv Mortensen - Cross ORD Post Doc
Doris Smith - Technical SEE Grantee
James Vail - Technical SEE Grantee
David Reif
Ann Richard
Maritta Wolf- Lockheed Martin Contractor
J
Figure 3
D. BOSC Reviews of the CTRP
To help guide the CTRP, ORD established a standing subcommittee of the ORD Board of
Scientific Counselors to provide review and advice to NCCT and the CTRP. To view all prior
BOSC reviews and ORD responses, please visit NCCT BOSC Reviews.
The BOSC first met in April 2005 to review the organization of NCCT, initial plans for
implementation, and progress of the early CTRP work. The panel commented very favorably on
the NCCT's early progress and the steps, processes, and etc. outlined to achieve its goals. The
composition of staff, plans for future hiring, establishment of working partnerships, and the
Center's strategic plan especially were highlighted. Several recommendations were made for
consideration. Two main recommendations were to (1) develop a formal implementation plan for
5
-------
EPA CompTox Research Program FY2009 to 2012
Final
the future, and (2) to develop communities of practices (CoPs) within the EPA, which could
serve as a networking function for interested scientists. ORD's Computational Toxicology
Research Program Implementation Plan (FY 2006 - 2008 was completed in April 2006, in time
for the next BOSC review. Two CoPs have been organized and are discussed later in this
document. A few other minor suggestions also were made, which were addressed in the formal
ORD response to the review.
On June 19 and 20, 2006, a second BOSC site visit was held to assess and evaluate NCCT
progress in executing the first implementation plan and incorporating prior BOSC
recommendations. In this review, it was noted that, in the 16 months of its existence, "NCCT
has made substantial progress in (1) establishing goals and priorities; (2) making connections
within and outside EPA to leverage staffs considerable modeling experience; (3) expanding its
capabilities in informatics; and (4) significant contributions to research and decision making
throughout the Agency." The BOSC also noted that "many of the recommendations made by the
BOSC during its first review have been acted on by NCCT." This review occurred just before
NCCT hired its first two scientists under ORD's new Title 42 authority, which brought needed
experience in informatics and systems biology to address several of the key recommendations
of the BOSC from this second review.
The third review by the BOSC occurred on December 17 and 18, 2007. In this review, the
BOSC noted that NCCT continued to make substantial progress in setting priorities and goals
and specifically acknowledged the "increased capabilities in bioinformatics through the funding
of two STAR centers and in informatics and systems biology through staff hires; expansion of its
technical approaches to other programs within the Agency; and formation of the Tox21
collaboration with the National Institute of Environmental Health Sciences (NIEHS) and the
National Human Genome Research Institute (NHGRI) as a complement to its ToxCast™ project.
They went on to recommend that (1) client offices participate in future reviews to ensure all
parties understand how NCCT's efforts can address the most relevant needs of the Agency; (2)
interactions with risk assessors in the Agency be enhanced, particularly related to how
ToxCast™ might be utilized by them; (3) complementary efforts in exposure prioritization be
undertaken; (4) the directions and milestones be detailed and applications of the virtual tissues
to risk assessment be clarified; and, (5) finally, a more precise definition of the database for
compilation and rigorous quantitative analysis of the ToxCast™ data. Overall, the review
considered the goals of the Center to be well described, very ambitious and innovative, and
important for the future of research at EPA.
The next BOSC review is scheduled for September 29 and 30, 2009, and will focus on the
products from the first CTRP implementation plan and the future directions outlined in this
second generation plan.
E. NRC Report and EPA's Strategic Plan for 21st Century Toxicology
Two significant activities have increased the visibility and importance of EPA's computational
toxicology research efforts over the past 2 years. The first is an NRC report, commissioned by
EPA that presented a vision of how toxicological evaluations should be conducted in the future.
The second activity, motivated by the NRC report, was development of an EPA strategic plan to
transform how the Agency addresses chemical toxicity.
6
-------
EPA CompTox Research Program FY2009 to 2012
Final
1. NRC Report on Toxicity Testing in the 21st Century
In 2007, the NRC released a report titled Toxicity Testing in the Twenty First Century: A Vision
and Strategy, which outlines a long-term vision for developing novel approaches to chemical
toxicity characterization and prediction. This vision addresses several concerns about the
current best practice methods for toxicity characterization that rely heavily on extensive animal
testing. These concerns are the desire to reduce the number of animals used in testing, to
reduce the overall cost and time required to characterize each chemical, and to increase the
level of mechanistic understanding of chemical toxicity.
The NRC report outlines an approach for toxicity determination in which each chemical first
would be characterized for a number of properties related to environmental distribution,
exposure risk, physico-chemical properties, and metabolism. These activities fall under the
heading "chemical characterization," and (with the exception of metabolism) exclude bioactivity
in the target organism. The second stage is "toxicity pathway characterization," in which a series
of cell-based and non-cell-based (in vitro) tests would be used to indicate which (if any) "toxicity
pathways" are activated by the test chemical. A major challenge posed by this approach is that
few such toxicity pathways currently are understood, and assays to probe these pathways
therefore, generally are lacking. Next, for a subset of chemicals, "targeted testing" would be
carried out to refine our understanding of the effects of triggering specific toxicity pathways.
Targeted testing might involve additional in vitro assays or a limited amount of in vivo animal
testing. A final phase is "dose response and extrapolation modeling," which would use new and
existing data and models to perform low-dose extrapolation, toxicokinetics, and exposure
estimation. All phases would have significant computer modeling components. The end result of
these studies would be a determination of the potential toxic effects (including mode and
mechanism of action) of a compound, as well as estimates of dose-response behavior.
There is a relatively simple argument as to why an in vitro-based approach should be able to
predict whole animal or human toxicity. The effect of a chemical ultimately is caused by direct or
indirect molecular interactions of the chemical with one or more cellular components. These
interactions can be receptor or enzyme binding, disruption of a lipid membrane, localized
production of free radicals, or non specific dephosphorylation. Nonetheless, if two chemicals
have the same biological interactions and have the same distribution and kinetics within an
organism, then the two chemicals should present the same bioactivity profiles and potential
toxic effects. This concept highlights a major benefit of the in vitro, mechanism-based approach;
it provides a way to extrapolate from one chemical to the next based on a set of relatively
inexpensive and quick biochemical or cellular assays. Achieving this vision, however, will take
many years because of a number of circumstances that are described in the NRC report. It is
noteworthy that the NRC report did not downplay how difficult the task of developing this new
approach to toxicity testing would be, proposing a timeline of 20 years, with annual investments
on the order of $100 million to fully achieve this vision. However, for initial chemical screening
and prioritization, a generalized set of assays for key biological targets and pathways can be
implemented successfully in a much shorter and less costly fashion.
2. EPA Strategic Plan for Evaluating the Toxicity of Chemicals
In response to the release of the NRC report, EPA established an intraagency workgroup, the
Future of Toxicity Testing Workgroup (FTTW), under the auspices of the Science Policy
Council. The FTTW includes representatives from across the Agency, including the regions and
program offices. It produced The U.S. Environmental Protection Agency's Strategic Plan for
Evaluating the Toxicity of Chemicals, which serves as a blueprint to ensure a leadership role for
7
-------
EPA CompTox Research Program FY2009 to 2012
Final
EPA in pursuing the directions and recommendations presented in the 2007 NRC report. The
strategy presents the Agency's vision of how to incorporate a new scientific paradigm and new
tools into toxicity testing and risk assessment practices with ever-decreasing reliance on
traditional approaches. The overall goal of the strategy is to provide the tools and approaches
necessary to move from a near exclusive use of animal tests for predicting human health effects
to a process that relies more heavily on in vitro assays, especially those using human cell lines.
The program envisioned in this strategy builds on the traditionally major components of the risk
assessment process (i.e., hazard identification, dose response, exposure assessment, risk
characterization) by overlaying a toxicity pathways approach on the
source—>fate->-transport—>-exposure—K)utcome continuum. Specific components include the
three integrative and interactive focal areas described below:
(1) Toxicity Pathway-Based Chemical Screening and Prioritization. This research will
focus on identifying toxicity pathways and deploying in vitro assays to characterize the ability
of chemicals to perturb those pathways, and further development and implementation of the
ToxCast™ concept to establish the predictive relationship of the new assays for identifying
adverse outcomes in humans or ecological populations.
(2) Toxicity Pathway-Based Risk Assessment. This element will focus on reducing key risk
assessment uncertainties currently associated with the extrapolation of data from animal
studies to humans, from high doses to relevant human exposures, and to different
population susceptibilities (e.g., children). The program will achieve these ends by
developing knowledgebases of toxicity pathways, toxicological responses, and information
on biological networks; constructing dynamic computational models of tissue biology that
link diverse data together to understand the progression of events from exposure to effect;
and demonstrating that this new vision of toxicity testing adequately predicts human risk
using case studies.
(3) Institutional Transition. Implementing major changes in toxicity testing of environmental
chemicals and incorporating new types of toxicity data into risk assessment will require
significant institutional changes in terms of how EPA transitions to the use of new types of
data and models; how EPA deploys resources necessary to implement the new toxicity
testing paradigm, such as hiring scientists with particular scientific expertise and training
existing scientific staff; and how EPA educates stakeholders and the public.
Although the workgroup identified a range of partners in this effort, and some planning on the
relative role of these partners has been developed, the specific areas of work to be conducted
and funded by EPA versus these partners, needs to be further assessed. Decisions on these
relative roles will have a significant impact on EPA resources required to implement the vision.
Regardless, the CTRP will play a central role in all three strategic goals from identifying and
conducting high-throughput screening on chemical libraries of interest, to developing systems-
level models of biology for application in risk assessment, to training program offices in the
understanding and use of the new technologies.
8
-------
EPA CompTox Research Program FY2009 to 2012
Final
F. Significant Accomplishments of the CTRP: FY2006 to 2008
1. Accomplishments of the NCCT
With the CTRP being in existence for nearly 6 years, and the NCCT for more than 4 of those, a
number of projects contained within the CTRP are now delivering important products,
information, and accomplishments to regulatory offices within the Agency, toxicologists in the
international community, and the regulated chemical industry. Details of such accomplishments
are excerpted below, and greater detail is available in the descriptions of the 9 NCCT-led
projects provided in Appendix IV.A., with additional information on the annual milestones and
projected impacts of the projects provided in summary format in Appendix IV.C.
A number of projects on informatics, chemical prioritization, and systems-level models of biology
are beginning to provide the foundation for high-throughput, decisionmaking tools that EPA
program offices can apply to chemical hazard screening and risk assessment. Descriptions of
these projects are presented below:
• ToxRefDB. A relational, electronic toxicity reference database (ToxRefDB) was developed in
partnership with the Office of Pesticide Programs (OPP), that contains results of more than 30
years and $2 billion worth of rat and mouse chronic, rat multigenerational, and rat and rabbit
developmental studies for more than 400 chemicals (Martin et al. 2009a ; Martin et ai. 2009b;
Knudsen et al. 2009). This relational database is enabling the Agency, for the first time, to
readily discern patterns of toxicity in the assays and to assess the value of the design of the
assays in assessing toxicity. It also will be invaluable in the interpretation of the HTS data
derived in the ToxCast™ Program. This database will be expanded to include developmental
neurotoxicity assays, and potentially, results from the Endocrine Disruptor Screening
Program, thus affording a one-stop shop for animal bioassay data.
• ACToR. The Aggregated Computational Toxicology Resource (ACToR) provides an Internet-
based portal of information on chemical structure, bioassay, and toxicology data for
environmental chemicals from 200 sources of public data for more than 500,000 chemicals
and a central integrated public resource for all DSSTox, ToxCast™, and ToxRefDB data
(Judson et al. 2009). ACToR was released to the public in the second quarter of FY2009 and
will undergo periodic updates over the next several years as we add data and functionalities,
and respond to user feedback.
• DSSTox. The Distributed Structure-Searchable Toxicity Database Network (DSSTox) was
updated with several high-interest chemical inventories, including ToxCast™ Phase I
chemicals, EPA high production volume chemicals, and the two primary public genomics
inventories (GEO and Array Express), an online structure-browser, and linkages and
coordination with internal and outside resources such as ACToR and PubChem (Williams et
al. 2009; Wlliams et al. 2009). DSSTox, in coordination with ACToR, is also responsible for
all chemical information registration and review for ToxCast™ and Tox21 (see below for a
more detailed explanation of Tox21) projects and is the source of high-quality chemical
structures for the Organization for Economic Co-Operation and Development (OECD) QSAR
Toolbox.
• ToxCast™. ToxCast™ is a major effort to evaluate the comprehensive use of HTS assays to
provide biological fingerprints of activity that can be used to predict adverse outcomes in
rodents and in humans (Dix etal. 2007). Wth the initiation of nine contracts in FY2007, the
CTRP began to generate a myriad of molecular in-vitro based data for the ToxCast™
program. Data collection for more than 300 chemicals from over 500 assays in Phase I began
in FY2007 and was completed in the second quarter of FY2009. An internal EPA workshop
9
-------
EPA CompTox Research Program FY2009 to 2012
Final
regarding ToxCast™ was held in March 2008, and a public ToxCast Data Analysis Summit
workshop was held in May 2009, with more than 30 national and international material
transfer agreements (MTA) analysis partners participating. Analyses are focusing along three
dimensions: (1) comparing bioactivity profiles across chemical classes; (2) correlating specific
assays or pathways with toxic phenotypes; and (3) correlating phenotypic syndromes with
bioactivity profiles. It is expected that analysis of the data will continue over several years,
both by EPA and external groups, as the compendium of data represents a truly unique and
innovative resource. A significant number of additional partners were brought into the
program in FY2008 and FY2009 via MTAs. Additional contract awards are anticipated to
augment the breadth of biological pathways contained in Phase II, which will launch in late
FY2009. A critical feature of Phase II is the plan to include drugs that have failed in human
clinical trials, thus providing human toxicity data on which to benchmark ToxCast™ bioactivity
profiles. One major pharmaceutical company, Pfizer, already has agreed to collaborate in this
regard, and the Health and Environmental Sciences Institute has adopted the concept as their
Emerging Issues Proposal for 2009, promoting the opportunity for a much wider group of
pharmaceutical companies to contribute failed drugs and clinical data. Thus, Phase II of
ToxCast™ will be generating HTS data on 700 additional chemicals, some with animal toxicity
information, clinical data, and additional data on human disease, susceptibility, and variability
that will contribute to the goal of expanding, verifying, and translating in vitro bioactivity into
predictions of potential toxicity.
• Tox21. To garner complementary expertise across the federal government to transform the
field of toxicology to a more predictive science, EPA signed a memorandum of understanding
(MOU) with the National Toxicology Program (NTP)/National Institute of Environmental Health
Sciences (NIEHS) and the National Institutes of Health Chemical Genomics Center (NCGC)/
National Institutes of Health Chemical Genome Research Institute (NHGRI) in February 2008
(Collins et al. 2008). This Tox21 consortium now has four active working groups identifying
chemicals, assays, informatic analyses, and targeted testing as plans proceed to have nearly
10,000 chemicals under study at the NCGC by the third quarter of FY2010 (Kavlock et al.
2009). Supported by interagency agreements between the NTP and EPA, this effort will
conduct 50 or more HTS assays on this enlarged chemical library every year for the next
several years.
• Cumulative Risk. Tools of computational toxicology for modeling exposure, effects and risk
have been used in integrating data for the cumulative risk assessments of cholinesterase-
inhibiting pesticides (organophosphates_and carbamates), particularly for modeling and
characterizing the confidence and uncertainties in the assessment. These methods have
been used to set safe exposure limits for these chemicals in the field and home, and are also
being used to help in the consideration of a cumulative risk assessment for the pyrethroid
pesticides.
• Communities of Practice (CoP). Though not projects or tools in the typical sense, NCCT
has formed several CoPs promoting the utilization of CTRP research in specific areas of
computational toxicology. Each CoP has a charter and an open membership policy and is co-
chaired by a member of NCCT. The CoPs operate via activities such as teleconferences,
face-to-face meetings, team rooms, and workshops. By bringing together members from
different parts of EPA, ORD, and the outside scientific, regulatory, and regulated community,
CoPs help to promote the adoption of common practices and ontologies, guide development
of common databases and software usage, aid in construction of training materials, provide
recommendations on efficiencies of relevant operations, and act as a public outreach
mechanism for ORD activities. To date, two CoPs have been established on Chemical
Prioritization and Exposure Science. Both of these are large, active groups that meet monthly
and have brought a wealth of ideas, interest, and international collaboration to various CTRP
projects.
10
-------
EPA CompTox Research Program FY2009 to 2012
Final
• International Science Forum on Computational Toxicology. In 2007, the CTRP hosted
EPA's annual Science Forum. This was the first EPA Science Forum held outside of
Washington, DC, and featured an international overview of the state of the science on
computational toxicology (Kavlock et al. 2008). More than 300 people attended this forum on
advances in computer sciences, molecular biology and chemistry, and systems models that
can be used to increase the efficiency and the effectiveness by which the hazards and risks of
environmental chemicals are determined. The forum surveyed the state of the art in many
areas of computational toxicology and identified key areas important to move the field
forward: proof-of-concept studies demonstrating the additional predictive power gained; more
researchers comfortable generating and working with high-throughput data and using it in
computational modeling; and regulatory authorities willing to embrace new approaches as
they gain scientific acceptance. Ideas from this forum and efforts bridging from it were helpful
in EPA's development of The U.S. Environmental Protection Agency's Strategic Plan for
Evaluating the Toxicity of Chemicals.
2. Accomplishments by the CTRP ORD Intramural Partners
The seven cross-ORD "new start" research projects initiated by the CTISC in FY2004 advanced
the field of computational toxicology and were important components in the first generation
CTRP implementation plan and the establishment of NCCT. Details on the organization, goals,
and cross-ORD participants for these projects were provided in the FY2006-2008 CTRP
Implementation Plan, and a comprehensive listing of publications from these projects is included
in Appendix IV.C. Descriptions of major accomplishments are provided below.
(1) Linkage of Exposure and Effects Using Genomics, Proteomics, and Metabolomics in
Small Fish Models. This project used a combination of whole organism end points;
genomic, proteomic, and metabonomic approaches and computational modeling to (a)
identify new molecular biomarkers of exposure to endocrine-disrupting compounds (EDCs)
representing several modes and mechanisms of action (MOA) and (b) link those biomarkers
to effects that are relevant for both diagnostic and predictive risk assessments using small
fish models (Villeneuve et al. 2007; Ankley et al. 2009). Data from the project have provided
the basis for several important predictive modeling efforts. For example, the development of
a graphical systems model focused on defining the HPG axis of small fish, which enables
consideration of the interactive nature of a perturbed system at multiple levels of biological
organization, ranging from changes in gene, protein and metabolite expression profiles to
effects in cells/tissues that directly influence reproductive success. A second modeling effort
involves development of a steady-state model for ovarian tissue to predict synthesis and
release of testosterone and estradiol. Results from the model were compared successfully
to data generated from the fathead minnow. Model-predicted concentrations of the two
steroids over time corresponded well with both baseline (control) data and information from
experiments in which estradiol synthesis was blocked by fadrozole. Modeling also has
focused on the consequences of perturbations in the HPG axis relative to effects in
individuals and populations. Here, a population model that employs a Leslie matrix in
conjunction with the logistic equation (to account for density dependence) was used to
translate laboratory toxicity information into prediction of population trajectories. In these
analyses, changes in steroid or vitellogenin concentrations in female fish first were related to
fecundity, and then, using this relationship, to population status in fish exposed to EDCs that
inhibit production of vitellogenin most notably compounds that depress steroid synthesis
(e.g., fadrozole, prochloraz, trenbolone). That analysis is unique in that it focuses on
biochemical end points, female steroids, and vitellogenin, which reflect both toxic MOA of
EDCs and have a functional relationship to reproductive success (formation of eggs). As
11
-------
EPA CompTox Research Program FY2009 to 2012
Final
such, within the overall systems framework for the project, this computational model can
serve as the basis via which genomic information can be linked quantitatively to responses
in populations.
(2) A Systems Approach to Characterizing and Predicting Thyroid Toxicity Using an
Amphibian Model. The main objective of this work was to develop a hypothalamic-pituitary-
thyroid (HPT) model that is capable of integrating data from different levels of biological
organization into a coherent system (Deqitz et al. 2005). A simulation model has been
developed to describe the thyroid axis of X. laevis tadpoles. Information pertaining to normal
baseline HPT-axis development was collected to compare to the perturbed system. Thyroid
and pituitary gland culture systems were developed to investigate the response of these
components in isolation from the HPT-axis feedback mechanisms within the animal. Genes
responsive to TSH were also measured in the thyroid gland and pituitary in vivo and in vitro
in response to chemical exposure and TSH stimulation. Genes involved in the T4 synthesis
pathway that robustly responded to TSH, such as NIS and thyroid peroxidase (TPO), were
not changed when challenged with the chemical alone. Other thyroid-specific genes, such
as thyroid transcription factor and thyroglobulin, were not TSH-responsive. Efforts were
made to obtain TPO activity from X. laevis, but the small amount of tissue contributed to our
inability to develop an assay. Therefore, porcine thyroid glands were used to isolate and
measure TPO activity. Twenty-four chemicals have been tested in the in vitro assay for their
capacity to inhibit TPO activity. Although the majority of the tested chemicals were negative,
several were identified as TPO inhibitors. These positives were tested in the thyroid gland
explant culture assays and the TH released into the culture media was measured by RIA.
One of the test chemicals, a mercaptobenzothiazol, inhibited T4 release from the thyroid
glands at concentrations that were not overtly toxic to the gland. Chemicals that test positive
in the TPO inhibition assay and the ex vivo thyroid gland TH release assay will be tested in
the abbreviated amphibian metamorphosis assay to determine activity in the HPT axis in
vivo. This will begin to provide information on the predictive ability of the in vitro and ex vivo
assays for identifying thyroid-disruptive chemicals.
(3) Risk Assessment of the Inflammogenic and Mutagenic Effects of Diesel Exhaust
Particulates: A Systems Biology Approach. This project utilized a systems approach to
developing and applying predictive computational models that quantitatively describe
relationships between the composition of DEP and its genotoxic and inflammogenic
potencies (Linak et al. 2007; Stevens et al. 2009). A significant accomplishment is derived
from the development of a prototype ESP sampler, which has resulted in a significant
improvement in DEP collection yield relative to conventional filter methods for particle
capture that were in use previously. These analyses have produced an unprecedented
physicochemical characterization of DEP, including XRF analysis of metal content and
GC/MS analysis of organic species, as well as determinations of OC/EC, particle size and
aerodynamic characteristics, etc. By design, this Phase generated inflammogenicand
mutagenic DEPs of varying composition to address the central hypothesis regarding the
inflammogenic and mutagenic potency of DEP. An extensive database has been generated
and currently is being analyzed for publication and use in model development.
(4) The Mechanistic Indicators of Childhood Asthma (MICA) Study: A Systems Biology
Approach to Improve the Predictive Value of Biomarkers for Assessing Exposure,
Effects, and Susceptibility in the Detroit Children's Health Study. This computational
toxicology effort integrated rodent and human research across the source-to-outcome
continuum to link gene expression with clinical outcomes and biomarkers of exposure, early
effect, and susceptibility for two broad classes of chemicals: (1) polycyclic aromatic
hydrocarbons and (2) metals (Heidenfelder et al. 2009; Edwards et al, in press). In
collaboration with Michigan State investigators, exposures of rodents to concentrated air
particulates were completed using state-of-the-art mobile air exposure chambers. Genomic
12
-------
EPA CompTox Research Program FY2009 to 2012
Final
analysis of rodent blood and lung tissues highlighted tissue-specific patterns of gene
expression related to airborne exposures. For the children's study (whose protocols were
approved by three separate internal review boards), a 20-page respiratory health
questionnaire was mailed to the parents of 6,883 children aged 7 to 12 years recruited
through the Henry Ford Health System. An indoor/outdoor MICA-Air component was added
using an innovative participant-based air sampling approach that included measurements of
nitrogen dioxide and selected volatile organic compounds. A subset (205) of these children
had clinical examinations including measurements of pulmonary function and exhaled nitric
oxide, collection of blood for both genetic and gene expression analysis, and nail and urine
samples collected. Our analysis strategy represents a true "systems" approach in that each
type of data is examined in the context of the broader MICA data set. This systems
approach affords more robust conclusions, because the predictive value of biomarkers from
particular data slices can be assessed for biological and statistical validity against the
diverse set of supporting data.
(5) An Approach to Using Toxicogenomic Data in Risk Assessment: Dibutyl Phthalate
Case Study. To address how genomic data may be used most effectively in risk
assessment, this project was initiated with the goals of developing an approach to using
toxicogenomic data in risk assessment and testing this approach in a case study.
Recognizing that genomic data type (e.g., species, organ, design, method) varies, the
approach included formulating questions that the toxicogenomic data may inform. Because
microarray data is often informative of the MOA of a chemical, the approach included an
assessment of the toxicogenomic dataset in conjunction with the toxicity dataset to relate
the affected end points (identified in the toxicity dataset evaluation) to the pathways
(identified in the toxicogenomic dataset evaluation) as a method for informing the MOA.
Dibutyl phthalate (DBP) was selected for the case study, focusing on the male reproductive
outcomes, because it has a relatively large and consistent genomic dataset, phenotypic
anchoring of certain gene expression data for these outcomes, and an ongoing Integrated
Risk Information System (IRIS) assessment. The case study team concluded (USEPA,
2009) that the "genomic dataset" should include all gene expression data (single gene,
global gene expression, protein,and RNA) in the evaluation, as these data taken together
provide a stronger basis for reproducibility of the global gene expression study findings. This
evaluation found that the gene level findings from the DBP genomic studies (i.e., microarray,
RT-PCR, protein expression) were highly consistent in both the identification of differentially
expressed genes (DEGs) and their direction of effect. This project also identified research
needs for toxicity and toxicogenomic studies for use in risk assessment. These include (1)
parallel study design characteristics with toxicogenomic studies (i.e., dose, timing of
exposure, organ/tissue evaluated) to obtain comparable toxicity and toxicogenomic studies
to aid our understanding of the linkage between gene expression changes and phenotypic
outcomes, (2) exposure time-course microarray data to develop a regulatory network model,
(3) generate TK data in relevant study (time, dose, and tissue) and obtain relevant internal
dose measure to derive best internal dose metric, and (4) multiple doses in microarray
studies in parallel with phenotypic anchoring.
(6) Development of Microbial Metagenomic Markers for Environmental Monitoring and
Risk Assessment. This project focused on the development of nonculture-based genomic
methods for environmental monitoring and risk assessment. The research focused on the
use of a microbial community genome (metagenome) approach to identify novel nucleic acid
sequence markers for fecal contamination and source identification (Shanks et al. 2006;
Santodominqo et al. 2007). The basic experimental design consisted of challenging genomic
microbial community DNA from different fecal samples in genome subtraction studies to
enrich for host specific microbial genes. Sequencing complete fecal metagenomes
generated redundant information, making difficult the selection of host-specific genes. To
13
-------
EPA CompTox Research Program FY2009 to 2012
Final
address this limitation, a novel approach called genome fragment enrichment was
developed to select for DNA fragments present in a specific fecal microbial community and
absent in other fecal communities. A patent disclosure was filed on this specific method. In
each case, hundreds of enriched DNA fragments were sequenced and assigned a putative
functional role. Although several assays were developed, these have to be further
evaluated, particularly to determine the geographic stability of these methods, not only in the
United States, but also in other parts of the world. To this end, we are working with
researchers from different regions in the United States and from countries such as Canada,
Brazil, Austria, Singapore, and Spain. Results of this research will overcome the current
limitation of assessing the microbial water quality by measuring bacterial densities, which
are both time consuming and do not provide information about the sources impacting a
watershed. Such information is necessary to implement adequate pollution control and
remediation practices.
(7) Simulating Metabolism of Xenobiotic Chemicals as a Predictor of Toxicity. The
MetaPath research project is a collaboration with OPP scientists developing a capability for
forecasting the metabolism of xenobiotic chemicals of EPA interest, to predict the most likely
formed chemical metabolites, and to interface that information with toxic effect models,
enabling prediction of parent chemical toxic potential and the identity of chemical
metabolites of equal or greater toxicity than the parent chemical (Kenneke 2006). Key
milestones include developing and expanding the in vivo and in vitro liver metabolism
database, especially for chemicals and transformation reactions underrepresented in the
current database; finalizing development of the searchable metabolism database and
continuing to populate existing databases with additional metabolism data; enhancing the
performance of the existing metabolic simulator by incorporating reliable metabolism data
and expansion of relevant transformation reactions; and conducting in vitro experimentation
to verify maps and metabolites forecasted by the metabolic simulator and evidence for
enhanced estrogenicity.
3. Accomplishments by STAR Grantees in the CTRP
The STAR grants and centers have been a critical component of the CTRP from its inception. A
brief summary of the accomplishments of the three existing STAR centers for this program is
provided below. More information and fuller descriptions of these centers are available in
Appendix IV.B.
NCER funded two STAR environmental bioinformatics centers as part of the CTRP in FY2006.
The Research Center for Environmental Bioinformatics and Computational Toxicology at the
University of Medicine and Dentistry of New Jersey (UMDNJ), Piscataway, NJ, and The
Carolina Environmental Bioinformatics Research Center at the University of North Carolina
(UNC), Chapel Hill, are operating as cooperative agreements and helping to facilitate the
application of bioinformatics tools and approaches to environmental health issues supported by
the CTRP.
To date, the UMDNJ center has made progress expanding the framework of the Food and Drug
Administration ArrayTrack to ebTrack, an integrated bioinformatics system for environmental
research and analysis enabling the integration, curation, management, first-level analysis, and
interpretation of environmental and toxicological data from diverse sources. Other major
accomplishments include the enhancement of Shape Signatures QSAR technology for chemical
hazard identification, metabolic engineering tools for identifying important pathways within the
overall hepatocyte metabolism, and computational procedures for quantifying the structure of
14
-------
EPA CompTox Research Program FY2009 to 2012
Final
molecular bionetworks via the S-space Network Identification Protocol and the Closed-Loop
Identification Protocol.
Major accomplishments of The Carolina Center for Environmental Bioinformatics include the
development and refinement of a mouse model of variation in genetic susceptibility relevant to
human populations, pathway modeling in genomic analysis, and new methods in QSAR
modeling relevant to toxicity. This work complements other work in the CTRP, utilizing the
unique strengths of the STAR center in genetics, toxicology, and statistical modeling. An early
outcome of this work included dissection of the genetic regulation of liver gene expression. In
addition, the center has refined expression quantitative trait locus (eQTL) analysis procedures.
These methods serve the larger goal of elucidating the underlying mechanisms of toxicity. The
center also has developed high-quality methods for testing biological pathway involvement in
toxicogenomics studies, and a novel, hierarchical, two-step approach to model chemical
structure for in vitro/in vivo toxicity data.
In FY2008, NCER funded a third center, through a cooperative agreement, The Carolina Center
for Computational Toxicology at UNC in Chapel Hill. This center is applying high-performance
computing techniques and resources to in silico multiscale modeling applications at the cellular,
organ, and system-wide levels. In its first year, the center began to implement and design
advanced mathematical approaches to modeling biological systems and biological-chemical
interactions represented in ToxCast™, as well as other datasets.
Another high priority for EPA is to understand the molecular and cellular processes that, when
perturbed, result in developmental toxicity. With a project start date of November 2009, NCER
responded to this need by funding the Texas-Indiana Virtual STAR Center: Data-Generating in
vitro and in silico Models of Developmental Toxicity in Embryonic Stem Cells and Zebrafish at
the University of Houston, the Texas A&M Institute for Genomic Medicine, and Indiana
University. This center will bridge the interface of in vitro data generation and in silico model
development to answer critical biological questions related to toxicity pathways important to
human development. This research should result in an improved predictive capacity for
estimating outcomes or risks associated with developmental exposure to environmental
toxicants.
G. Retrospective Summary of the CTRP and NCCT
The first 5 years of the CTRP and NCCT have seen a great deal of progress and
accomplishment. These accomplishments have come from across ORD, the STAR grantees
and, increasingly, from NCCT, as the center becomes fully staffed and projects mature. As the
plan from the first implementation comes to an end, the CTRP is poised to carry out the vision of
the NRC report on Toxicity Testing in the Twenty First Century: A Vision and Strategy, and The
U.S. Environmental Protection Agency's Strategic Plan for Evaluating the Toxicity of Chemicals.
This CTRP implementation plan for FY2009 to 2012 will go on to explain revisions to the
program and specific projects that will be management priorities into the future.
15
-------
EPA CompTox Research Program FY2009 to 2012
Final
II. Revision of the CTRP for FY2009 to 2012
A. Maturation of the Program
The accomplishments of the CTRP noted above have enabled the program to evolve beyond
the early focus on hazard identification and chemical prioritization into broader areas of risk
assessment. In doing so, a number of research activities are becoming increasingly intertwined.
For example, results of the ToxCast™ program are providing data for models being developed
in the virtual tissue programs, and, in turn, the virtual tissue programs are guiding targeted
testing needs for other parts of the program. As a result, a blending of activities across the 3
LTGs in the first implementation plan has happened. Although the 3 LTGs served us well in the
initial years of the program, we have reduced them in the current plan to a single goal, namely,
providing high-throughput decision support tools for screening and assessing chemical
exposure, hazard and risk. Following The U.S. Environmental Protection Agency's Strategic
Plan for Evaluating the Toxicity of Chemicals, we will see increased emphasis on the use of new
data types being generated in the CTRP in quantitative risk assessment and greater
involvement in aspects of high-throughput exposure models and in analysis of results coming
out of high-throughput bioactivity profiling. As noted in other places, discussions already are well
underway among NCCT, NHEERL, NERL, and NCEA on the shape of such an expansion. This
expansion will incorporate progress from ToxCast™ and ExpoCast™, as well as the v-Liver™
and v-Embryo™ projects, into informatics tools and databases that can be used for both
research and regulatory applications. For example, in FY2011 we expect ToxCast™ will be
winding down Phase II of its three-part development program, providing in vitro and in vivo
toxicity data on a total of 1,000 compounds across hundreds of molecular targets and biological
pathways. Over this same time, the Tox21 consortium will be building a chemical screening
library of as many as 10,000 chemicals, conducting approximately one assay per week on this
library, and providing additional in vitro bioactivity data. Within the ACToR database and Web
site and in the analysis application ToxMiner, all of this bioactivity data will be merged with
chemical information and the exposure data being curated from the ExpoCast™ project. At this
stage in FY2011, steps will be taken to provide EPA's program offices decision analysis tools
that incorporate hazard and exposure data for prioritization of further chemical testing. Prior to
that time, the CTRP will be hosting periodic training courses on the new science and tools for
program office and regional personnel. We envision this decision analysis tool being part of the
ToxMiner application. In addition to statistically based predictive toxicology tools, ToxMiner will
be able to reference the mapping of ToxCast™ assays to biological pathways curated from
sources such as the Kyoto Encyclopedia of Genes and Genomes into a computable format for
use in ToxMiner.
Depending on the precise pace of progress in ToxCast™ and ExpoCast™ and the development
of screening and prioritization tools, additional resources could be freed up to support other
components of the CTRP in FY2011 and FY2012. However, chemical prioritization will remain a
priority area until it reaches the stated objectives. A natural extension of the chemical
prioritization projects is using this data for systems modeling in the v-Liver™ and v-Embryo™
projects. It is these systems models that will help fully utilize the toxicity and exposure pathway
information coming out of ToxCast™ and ExpoCast™ in next-generation, higher throughput risk
assessments. In conjunction with Office for Prevention, Pesticides and Toxic Substances
(OPPTS), we also have begun an examination of the feasibility of using HTS tools to evaluate
the potential hazards of nanomaterials. This effort will begin with carbon nanotubes and
nanomaterials being used in the OECD Working Party on Manufactured Nanomaterials and
include materials submitted to the EPA via the program Premanufacturing Notification of the
16
-------
EPA CompTox Research Program FY2009 to 2012
Final
Toxics Substances and Control Act. We expect the application of bioactivity profiling of
nanomaterials to become an increasingly important component of the CTRP as we gain
experience with their handling and screening results.
Relative to the STAR program, the number of funded centers will reach four or five (2 on
environmental bioinformatics, 1 on computational toxicology, and 1 or 2 on virtual tissues) by
FY2010 and consideration will need to be given to the second-generation centers as the first
ones to be awarded begin to reach the conclusion of their funding cycle.
For historical purposes, the following is provided as transition information relative to the original
3 LTGs structure, with notation on the degree of emphasis the research activities that will be
carried over into the new implementation plan. Much of the increased activity is predicated on
continual and increasing support of the CTRP program, such as is being witnessed in the
FY2010 budget process. The following section describes research areas of increased and
decreased effort relative to the 3 FY2006 to FY2008 LTGs.
Long-Term Goal 1. EPA risk assessors use improved methods and tools to better understand
and describe linkages across the source-to-outcome paradigm. Work was directed toward
computational models and modeling systems that represented comprehensive descriptions of
the underlying biology of adverse impacts caused by exposure to environmental agents. The
whole-systems biology modeling approach was to develop a range of models, from those
describing pharmacodynamic connections between exposure and effects to those describing
complex endogenous pathways and the perturbations in such pathways resulting from
environmental exposures. Also, ways to incorporate and use "omics" information in these
models was explored. Finally, attempts were made at formulating models of common but
complex disease processes that then are exacerbated by exposures to exogenous substances
and stressors through the development of virtual organ models, the first being the virtual liver,
and the second the virtual embryo.
• Increase efforts to develop virtual models of tissues (liver and embryo) that link across levels
of biological organization from molecular-to cellular-to tissue-level responses
• Increase coordination of efforts across NCCT, NERL, NHEERL, and NCEA to ensure models
of aspects of the source-to-outcome paradigm can be integrated and scaled to meet the
increasing needs of chemical evaluations
• Decrease research efforts related to the validation and acceptance of PBPK models
• Decrease, after FY2009, linkage of exposure and effects using genomics, proteomics, and
metabonomics in small fish models.
Long-Term Goal 2. EPA program offices use advanced hazard characterization tools to
prioritize and screen chemicals for toxicological evaluation. Molecular biological tools were
employed to develop fingerprints of biological activity of chemicals of concern to EPA.
Computational models were applied to the fingerprints to derive associations with classical
measures of toxicity derived from animal studies so that predictive models were developed,
leading to more efficient testing paradigms and reduction in uncertainties in interspecies
extrapolation. Proof-of-concept demonstration of ToxCast™, the forecasting tool, is scheduled
to provide a number of EPA program offices with an extremely useful tool to improve the
efficiency and effectiveness of hazard identification and risk assessment methodologies. There
were also new and innovative ways developed to assimilate, evaluate, and use the myriad of
data assorted with molecular and chemical information. Increasingly integrated chemical-
biological effects databases are intended to spur new capabilities for data mining and chemical
categorization in conjunction with HTS data. Development of an advanced computational
17
-------
EPA CompTox Research Program FY2009 to 2012
Final
chemistry method also provided in silico means to predict complex interactions of environmental
chemicals with biochemical receptors that then can lead to adverse effects.
• Increase support for Phase II of ToxCast™. The additional chemicals, which could total 1,000,
will include a greater number of pesticides, pesticidal inerts, antimicrobials, industrial
chemicals, water contaminants, and failed drug candidates. Phase II should be winding down
sometime in FY2011, depending on the exact pace of progress in FY2009 and FY2010.
• Increase development of methods to analyze data and relate the results from the ToxCast™
studies to potential for hazard and risk from realistic human exposure levels
• Increase interactions between NCCT and NHEERL on identification of additional critical
toxicity pathways to include in the chemical prioritization program
• Increase in conjunction with NERL, development of ExpoCast™, the exposure component of
a chemical prioritization process
• Increase cross-governmental collaborations to employ quantitative HTS assays to predict
human toxicity by engagement with relevant components of the NTP/NIEHS, NCGC/NHGRI,
and FDA in the Tox21 program.
Long-Term Goal 3. EPA risk assessors and regulators use new models based on the latest
science to reduce uncertainties in dose-response assessment, cross-species extrapolation, and
quantitative risk assessment. The intent of this goal was to develop additional key modules for
computational models of biological processes relevant to the induction of toxicity for high-priority
environmental chemicals. These modules would help assess the interaction of exposure to
environmental chemicals with other processes, such as underlying disease and concomitant
intake of pharmacological agents. As a result, EPA will be less reliant on default assumptions
for risk assessment and better able to accurately characterize the true uncertainty associated
with risk predictions for various chemical classes (e.g., EDCs) under conditions more relevant to
actual exposures and lifestyles
• Decrease chemical specific efforts vis a vis tools that have greater generic applicability
• Increase analysis of the resulting HTS data to (1) provide MOA information to specific risk
assessments being conducted by EPA, (2) provide rationale for grouping of cumulative risk
assessments based on toxicity pathways, (3) design a higher throughput risk assessment
approach for chemicals based on exposure potential and perturbation of toxicity pathways;
and (4) develop methods for analyzing and quantifying the uncertainty in dose-response
model predictions
These modifications to levels of effort and emphases are consonant with distillation of the 3
LTGs from the FY2006 to 2008 implementation plan into a single goal- Providing High-
Throughput Decision Support Tools for Screening and Assessing Chemical Exposure, Hazard
and Risk. This more efficiently will support the use of new data types being generated in the
CTRP in quantitative risk assessments, as well as provide high-throughput hazard and
exposure data and models for screening. Because the majority of the HTS data is being
generated for human molecular targets and pathways, this will support a transition from the
current dependence on animal-based toxicology. The combination of clinical data from
pharmaceutical partners, expanding efforts to bring in human data on exposure, and multiscale
systems models should make it possible to improve both the pace and quality of risk
assessments.
18
-------
EPA CompTox Research Program FY2009 to 2012
Final
B. CTRP Integration Across ORD Laboratories and Centers
Because of the relatively small size of NCCT and the ambitious nature of the CTRP mission, a
key part of the process of advancing the science involves developing partnerships that are both
within and external to ORD, to best leverage resources committed to the effort. Within ORD the
majority of health, ecological, and risk assessment research is contained within a variety of
Multi-Year Plans (MYPs) that incorporate efforts from multiple laboratories and centers and are
coordinated by National Program Directors (NPDs). In the case of the CTRP, the Director of
NCCT also leads the ORD Computational Toxicology Program. At present, there is an ongoing
active dialogue between ORD laboratories and centers, and relevant NPDs regarding the future
directions of the CTRP and other programs. To move beyond dialogue, NCCT has hosted or is
hosting, rotational fellows from across ORD (NERL, NHEERL and NCEA to date) that spend 4
or more months working with NCCT scientists on CTRP projects.
In response to the Administrator's priorities, a pilot effort is being considered by ORD that
addresses problems of broad national significance through highly integrated multidisciplinary
research efforts. In the last several months, NCCT, NERL, NHEERL, NRMRL, and NCEA have
been exploring opportunities for greater synergy in execution of the CTRP, and each laboratory
and center is actively developing implementation plans for such an effort. This integration will
include aspects of the Human Health Research Program MYP and the Safe Pesticides Safe
Products MYP. The challenge will be to ensure these plans are adequately integrated and the
outputs are suitably ambitious in nature. As EPA has just released its Strategic Plan for the
Evaluation of the Toxicity of Chemicals (U.S. EPA, 2009^ there is a window of opportunity to
continue this momentum within the CTRP, while also expanding efforts to include critical
aspects of exposure assessment (NERL), toxicity pathway coverage and targeted testing
(NHEERL), life cycle analysis of chemical use (NRMRL), and quantitative risk assessment
(NCEA). If successful, this integration of efforts across multiple laboratories, centers, and
research programs could meld the numerous ongoing ORD research efforts in computational
toxicology into a more functionally integrated multidisciplinary research program.
Figure 4 shows the stages in the source-to-outcome continuum on the horizontal axis, existing
components of ORD research and development infrastructure needed to support risk
assessment on the vertical axis, and products and tools for decision support on the far right. At
present, many of these efforts are not integrated and, in some cases, not fully compatible; a
major emphasis in the near future will be to bring more coordination and integration to these
currently diverse efforts. To overcome these challenges, integrated cross-disciplinary teams
involving exposure scientists, effects researchers, risk assessors, risk managers, and
computational modelers will need to be further developed.
19
-------
EPA CompTox Research Program FY2009 to 2012
Final
ORD's Computational Toolbo)
Risk Assessment Tools
Quantitative
Models
Populations
Individuals
EFS Exposure (e.g. SHEDS) PBPK (e.g. ERDEM)
Environmental/lndividual/Organ KB
Molecular/Cellular/Tissue KB
BBDRs
Prioritization
Tools
ExpoCast
ToxCast
MetaPath QSAR ToxMiner
ACToR
HEDS CHAD DSSTox HERO TTDB ToxRefDB
Ambient
Monitoring
Personal
Monitoring
Biomonitoring 1
Molecular Epidemiology (Bioindicators)
Figure 4
Understanding of complex, interrelated environmental stressors and potential impacts on human
health has grown tremendously in recent years. Basic and clinical sciences, however, have
significantly outpaced risk assessment science. Insights into health and disease exist, but it is
unclear how to incorporate this information into risk assessment with current methodologies.
Risk assessment will have to address a fundamental paradigm shift from a reliance on animal
toxicology data derived primarily from rodent bioassays. The need for this shift is made more
immediate by the challenges of applying new types of data stemming from advances in
computational toxicology and the huge volume of data that will be generated from the European
Union's Registration, Evaluation, and Authorization of Chemicals Program. High-throughput
data will need to be translated into knowledge to support science-based decisions in risk
assessment. The areas where this knowledge is expected to have an impact are defining
toxicity pathways and informing risk assessors about interpretation of multiple MOAs for toxicity
and providing insight into human variability in key pathways and human susceptibility. The
impacts of this information will be qualitative and quantitative. For this type of information to be
incorporated quantitatively, numerous challenges exist, not the least of which is extrapolation
from in vitro test systems to in vivo human health outcomes. However, the challenges are not
insurmountable if ORD, all of EPA, and key extramural partners work together in a coordinated,
multidisciplinary fashion. The initial impacts of this new paradigm will probably be seen in cases
of chemicals lacking significant datasets but for which toxicity predictions or rankings can be
developed from HTS data. These results may be used to derive estimates of relevant potency to
chemicals that have much larger databases and affect the same toxicity pathways. In addition,
challenges in extrapolating from effective concentration in vitro will require additional
considerations in the development of environmental exposure estimates. Examples of promising
20
-------
EPA CompTox Research Program FY2009 to 2012
Final
CTRP efforts with antimicrobials and pesticidal inerts; EDSP compounds; and industrial
chemicals, such as phthalates and perfluorinated chemicals; as well as with hepatocarcinogens
and teratogens are all underway. Results from these studies have been presented and are
being published in the peer-reviewed literature and incorporated into EPA databases and
decision support software tools.
The CTRP will continue to provide critical research components, work on integrating these
efforts across ORD, and facilitate the institutional transition necessary to ensure that these tools
integrated by EPA programs and regions. This includes quantitative and mechanistic
experimental data (e.g., ToxCast™) that are useful for chemical prioritization but also supports
systems models that could be useful for quantitative risk assessments. Data generated from
experimental systems can be used to define toxicity pathways and link them to adverse events
via the MOA. This definition of adversity, versus toxicity pathways and key events will be used
to identify the critical HTS tools for predicting disease outcomes in target populations. Future
efforts also will need to focus on the translation of the computational and high-throughput
methods into information that can be used in risk assessment. This is most apparent with
respect to actual use in quantitative risk assessment (i.e., being able to use in vitro or
computational methods to develop a point of departure for an IRIS assessment or other type of
risk assessment) but is also relevant to prioritization (i.e. how does the relative potency
information derived from the in vitro assays or models compare to in vivo potency) if information
from screening assays is going to be used to drive further testing decisions. Key issues in
developing the next generation of risk assessments include
• defining adversity—before these methods can be used in risk assessment, it will be vital to
understand how "perturbations of toxicity pathways" relate to adverse effects that are of
concern for human or ecological health; and
• development of decision methods and criteria for using high-throughput data in risk
assessment.
The large interdisciplinary projects required to meet the goals of this program are dependent on
well-integrated data repositories such as ACToR. These databases not only provide inputs for
empirical models for prioritization and hazard identification but also access to well-structured
data required for data mining needs to define and evaluate toxicity pathways and dose-
response modeling needs. As hazard identification and risk characterization tools are
developed, a gradual shift from reliance on empirical models supported by limited mechanistic
information to incorporation of detailed MOA for predicting hazard can proceed. The first step in
this process will be the development of prioritization tools out of ToxCast™, which incorporate
toxicity pathway information to identify chemicals of most concern and to direct more detailed
testing and modeling toward chemicals and toxicity pathways of highest importance. As
computational models are developed that relate the perturbation of a toxicity pathway
quantitatively to an adverse outcome (aided by development of virtual tissues), it should
become possible to not only prioritize but also screen for chemicals where the evidence is
convincing of either safety or toxicity. The toxicity pathways and MOAs for those chemicals
where evidence is inconclusive would then become the subject of further experimental study.
Eventually, quantitative prediction of risk from HTS data could be derived based on evaluation
of toxicity pathway predictions from systems models of tissues and organs (e.g., v-Liver™ and
v-Embryo™) and, perhaps, even simulated target populations.
A major step in higher throughput modeling and prediction of exposure will come from the CTRP
ExpoCast™ project, a collaboration of NERL and NCCT focused on providing Biologically
Relevant Exposure Science for 21st Century Toxicity Testing. The ExpoCast™ project is
described in more detail in the project plans (see Appendix IV.A). It should be noted that
21
-------
EPA CompTox Research Program FY2009 to 2012 Final
preexisting and other NERL research, databases, and models will be critical to the success of
ExpoCast™ and the broader goals, of ORD's CTRP. As ExpoCast™, ToxCast™, and other
modeling efforts succeed, it will be possible to prioritize testing and monitoring of large numbers
of chemicals and other environmental contaminants. These prioritizations would be feasible
because they would be based on hazard (H) and exposure (E) predictions of calculable and
reasonable uncertainty that were derived solely from in silico and in vitro data (see Figure 5).
The Future State: Using Hazard and Exposure
Predictions To Prioritize Testing and Monitoring
HE
HE
"he HE
High Exposure Potential l-E \ Low Exposure Potential
T oxCast
Low
Hazard
Prediction
l-E
l-E
l-E
l-E
Lower Priority for
Testing and Monitoring
l-E ToxCast Targets
H
H
H
^ToxCast Hazard Prediction
l-E.
Low Priority for
Bioactivity Profiling
Intelligent, Targeted Testing
Human Biomonitoring
Figure 5
As suggested by the NRC report "Toxicity Testing in the 21st Century: A Vision and a Strategy"
and The U.S. Environmental Protection Agency's Strategic Plan for Evaluating the Toxicity of
Chemicals, the majority of current CTRP efforts are centered on advancing toxicity testing for
assessing human health effects of environmental agents. However, under environmental
legislative mandates (e.g., the Toxic Substances Control Act; the Federal Insecticide, Fungicide,
and Rodenticide Act; and the Clean Water Act), most EPA programs regulate compounds to
ensure both environmental and human health risks are properly managed. Statutory language
and resulting policy typically require decisions for chemicals that encompass environmental and
human health risks such that the CTRP eventually also will need to develop higher throughput
and computational approaches for ecotoxicology and risk assessment. Notable progress has
been made in the previous CTRP implementation plan and within other ORD research programs
on the development and use of toxicity pathway models, toxicology knowledgebases (e.g.,
ECOTOX), and systems biology models (e.g., small fish model) in the field of environmental
science. In follow-up to the current CTRP implementation plan for FY2009 to 2012,
opportunities will exist to bring together relevant disciplines, data, and models across both
human health and environmental risk assessment applications. As noted previously, we will be
using existing resources and projects to leverage an increased attention to ecologically relevant
22
-------
EPA CompTox Research Program FY2009 to 2012
Final
areas; however, given the challenges of developing a system for relevance to human health
assessment, the near-term goals of the CTRP will have to remain largely focused on that area.
It is expected that future versions of the CTRP and related implementation plans will
accommodate further progress in ecotoxicology that can be incorporated and merged with the
efforts relating to human health in this current CTRP plan.
C. Regional and Program Interactions
The CTRP is working closely with program offices, including the OPPTS, and the Office of
Water (OW), to create informatics tools and databases of use toward the goals of both research
and regulation. Within the past year, full-day sessions have been spent briefing senior
management in these and other program and regional offices on the activities of the CTRP. This
has resulted in shared participation, input, and authorship between program and CTRP staff in
planning, research projects, scientific publications, and promulgation of regulatory decisions.
Many examples of collaborations with program offices exist, including DSSTox, ACToR,
ToxRefDB, and ToxCast™. Regional input has been more limited; briefings on the CTRP have
been given to specific regions (e.g., 6 and 7), and to the regional risk assessors group.
However, additional dialogue is needed as products begin to emerge from the program that
could be of use to the regions, especially in relation to the Toxics Release Inventory and
Superfund programs.
In moving forward over the next several years, the key points of engagement with the programs
will be their use of the DSSTox, ToxRefDB, and ACToR databases as sources of chemical,
toxicity and exposure information, and predictions and prioritizations based on results from
ToxCast™, ExpoCast™, and the Virtual Tissues projects. More detailed descriptions of what
these engagements will be are provided in the individual project descriptions in Appendix IV.A.
Some specific examples include the continued co-development of ToxRefDB to include
additional guideline test results, as well as to be positioned to record the output of the tier 1
endocrine disruptor screening batteries.
OPPTS-OPP. Driven by its needs to assess the effects of pesticidal inerts and antimicrobial
agents, both of which suffer from limited data availability; OPP has been a strong proponent of
this new approach to toxicity testing. OPP has defined a Strategic Direction for New Pesticide
Testing and Assessment Approaches focused on developing and evaluating new technologies
in molecular, cellular, and computational sciences to supplement or replace more traditional
methods of toxicity testing and risk assessment. This integrated approach to testing and
assessment is moving toward a new paradigm where in vivo (animal) testing is targeted to the
most likely hazards of concern. As defined by OPP, the path forward will include close
collaborations with the CTRP to predict chemical toxicity and exposure through application of
efficient and effective screening tools including new in vitro assays that rapidly provide biological
profiles of the toxicological potential of chemicals (i.e., ToxCast™). Exposure and biomonitoring
data also will be critical to interpreting toxicity data and evaluating the effectiveness of the new
testing and assessment paradigm. Over the next 5 years, OPP plans to enhance its integrated
approach to testing and assessment to better determine what toxicity data are needed to further
refine risk assessments for chemicals that do not have extensive toxicity information (e.g., inert
ingredients, certain antimicrobial and biochemical pesticides, metabolites, and degradates of
pesticide active ingredients). Over the next 10 to 15 years, as experience is gained and as
understanding of toxicity pathways increases, an enhanced integrated testing and assessment
approach will be implemented for all pesticides, including conventional agricultural pesticides.
This approach will fully integrate hazard and exposure data, along with advanced systems
modeling based on new in vitro data and an understanding of toxicity pathways to better predict
23
-------
EPA CompTox Research Program FY2009 to 2012
Final
risks and to determine what additional data are necessary (i.e., virtual tissues). The key goals of
integrated approaches to testing and assessment are to
• improve our ability to set priorities for what data to require,
• ensure that the data requirements are focused on the right issues, and
• efficiently reach the end result of effective risk assessment.
This approach would provide the ability to focus testing on pesticide chemicals and the effects
that most likely could result in harm. As a result, testing would
• use fewer animals,
• take less time,
• be less expensive in data generation and review, and
• explore a broader range of potential adverse effects.
These goals and approach are wholly consistent with the goals and approach of the CTRP, and
represent the close working relationship between OPP and ORD in developing these tools for
integrated approaches to testing and assessment.
CTRP researchers are working with OPP and other parts of OPPTS, and with the OECD to
explore and evaluate regulatory application of molecular screening assays in relation to chemical
testing guidelines. The goals of this effort are to provide tools for (1) improving the understanding
of mechanisms of toxicity; (2) identifying biomarkers of toxicity and exposure; (3) reducing
uncertainty in grouping of chemicals for assessments, interspecies extrapolation, effects on
susceptible populations, etc.; and (4) providing alternative methods for chemical screening,
hazard identification, and characterization. Also, the CTRP and OPP are working with the OECD,
on reducing and refining current animal testing guidelines using the ToxRefDB database and
other tools (e.g., Extended One-Generation Reproductive Toxicity Test Guideline).
OPPTS-OPPT. For the Office of Pollution Prevention and Toxics (OPPT) there will be several
key points of intersection with the CTRP over the next several years. These include using HTS
technologies and bioactivity profiling to lend biologically based support to strengthen and
potential revise of chemical categories currently in the new and existing chemicals programs, to
evaluate the relative hazard of chemicals being evaluated by the Design for the Environment
Program, for providing toxicity pathway data on specific groups of chemicals of high concern
(e.g., the perflourinates), and for evaluation of the feasibility of HTS approaches to
characterization of the bioactivity profiles of manufactured nanoparticles. On the exposure side,
the CTRP expects to be engaged with exposure efforts within OPPT, such as the Inventory
Update Rule, as procedures are developed for broad scale predictions of exposure potential
across the life cycles of chemicals. CTRP researchers are working with OPPT and with the
OECD to access and incorporate CTRP tools (e.g., DSSTox, ToxRefDB, ER QSAR models) into
the OECD QSAR Application Toolbox. The toolbox is an international effort to provide tools for
grouping chemicals, based on the understanding of mechanisms of toxicity, to extrapolate
properties and effects from tested chemicals to untested chemicals.
OPPTS-OSCP. For more than a decade the Office of Science Coordination and Policy (OSCP)
has led EPA's work to fulfill mandates contained within the Food Quality Protection Act of 1996
to identify chemicals that can interfere with the function of natural hormones (e.g., endocrine
disruptors). This is an MOA that may result in significant adverse consequences in developing
organisms (e.g., embryo, fetus, neonates, and children) should they be exposed to levels
sufficient to cause perturbations of their endocrine systems. Animal models have shown that the
levels that impact developing organisms can be much lower than those impacting adults, and
that exposures during development can lead to adverse effects not seen until adulthood. The
Agency has adopted a tier 1 screening battery that is designed to detect whether a chemical
24
-------
EPA CompTox Research Program FY2009 to 2012
Final
substance interacts with the estrogen, androgen, or thyroid hormonal systems. Combinations of
in vitro and in vivo assays are used to provide complementary measurements that detect the
endocrine disrupting potential of a chemical. Given the rapid advances in computational and
molecular sciences, discussions are underway with OPPTS on the next generation of tools that
could be more efficiently applied to the large number of chemicals of potential concern for ability
to disrupt the function of the endocrine system. Included in these discussions are the integrative
potential of contributions from CTRP in HTS of chemicals and providing informatics and analysis
solutions and tools; of the Human Health Research Program in conducting targeted follow-up
testing and exposure analysis; of the Endocrine Disruptors Research Program in understanding
MOAs, developing methods for assessing cumulative risks, and improving the ability to
extrapolate results across species; of the Human Health Risk Assessment Program in
assessing the contributions of multiple exposures (e.g., chemicals with common or different
modes of endocrine action) across critical life stages; and of the Risk Management Research
Program in providing information on chemical life cycles that pose the greatest potential
exposure and risk and the development of tools for greener chemicals or processes and other
mitigation strategies.
The CTRP has identified a large collection of nearly 10,000 chemicals that are of high priority for
EPA program offices. Using available funding, a subset of 700 will be used in Phase II of
ToxCast™ (cost per chemical of ~$20k). Going beyond the screening of these chemicals in the
full suite of ToxCast™ assays (>500), ORD has the capacity under existing contracts to acquire
the entire collection of chemicals and to screen them in a subset of ToxCast™ assays that cover
a broad spectrum of endocrine-related activities. The cost of such HTS assays would be on the
order of $1 k to 4k per chemical, depending on precisely which subset of ToxCast™ assays were
included. The use of HTS assays for receptor binding and transcriptional activation and QSARs
for these end points to obtain empirical or predicted information on chemicals for which no data
were available could provide an ordered list of which chemicals have the highest potential to
interact with the estrogen, androgen, and thyroid receptors. Along with exposure information,
this prioritized list then could be used to guide other effects research, exposure analyses, and
assessment and management activities in the various parts of ORD. The results could be used
to select chemicals for entry into subsequent phases of EPA's EDSP.
OW. For OW, the CTRP expects to be engaged with the assessment of the bioactivity of
chemicals on the Candidate Contaminant List (CCL) and the derivation of subsequent CCL lists.
In the course of identifying and assessing CCL compounds, the OW has to collect and analyze
hazard and exposure data on thousands of the same compounds addressed in ACToR,
DSSTox, and other CTRP databases. In future iterations of this process, the CTRP will be able
to assist the OWwith data curation and analysis and provide a wealth of new hazard and
exposure data that is computable, as well as prioritization tools designed to use this data.
D. Priority Areas for CTRP Management
1. Toxicity Predictions and Chemical Prioritizations Incorporating Exposure
With the publication of the first predictive bioactivity signatures from Phase I and the initial proof
of concept of the ToxCast™ program, we will have hazard predictions that can be incorporated
into prioritizing chemicals for further screening and testing. Phase II, scheduled for launch in
later 2009, will explore greater diversity of chemical structures and classes to evaluate the
robustness of the signatures identified in Phase I. As indicated in Figure 5, having viable
exposure predictions or estimations also will be critical to the envisioned prioritization scheme.
ExpoCast™ will provide an overarching framework for the science required to characterize
25
-------
EPA CompTox Research Program FY2009 to 2012
Final
biologically relevant exposure and can, thus, inform chemical prioritization by linking information
on potential toxicity of environmental chemicals to real-world health outcomes. NCCT
management will support continued collaborations with NERL and other critical partners within
EPA and externally to improve accessibility to EPA human exposure data and create a
consolidated EPA exposure database focused on measured concentrations in biological media.
As NCCT pushes ahead into Phase II of ToxCast™, expanding the compounds in the screening
program to include nanomaterials and chemicals with demonstrated human toxicities (e.g.,
failed pharmaceuticals), it will coordinate these efforts with ExpoCast™ to maximize utility of
datasets to develop predictive models and decision support software for chemical prioritizations.
2. Strengthening Cross-ORD Collaborations
Given the broad nature of the challenges facing computational toxicology, the CTRP must
engage collaborative partners across ORD to be successful. Cross-ORD collaborations have
been a part of the CTRP from its inception and will continue to be a dominant feature of the
program. Numerous collaborations from previous years will carry forward, including linkages at
the management level, such as the MOUs with NHEERL and NERL to provide NCCT with
administrative support functions for funds control, extramural management, quality assurance,
and information management. Key research partnerships have developed between NCCT and
the rest of ORD. In addition, NCCT continues to advise NCER on the formulation of ideas for
new computational toxicology RFAs, providing suggestions for scientific peer reviewers, serving
on relevancy reviews as appropriate and collaborating with the cooperative research partners of
the STAR grants program. ORD's multiyear planning process provides another opportunity for
linkage between the CTRP and related research efforts. Each of the MYPs is led by a National
Program Director with support from staff of the laboratories and centers. NCCT participates
actively in four of the MYP teams. Three contain similar research activities for screening and
prioritizing chemicals, (i.e., Endocrine Disrupting Chemicals [EDCs; LTG III], Safe
Pesticides/Safe Products [SP2; LTG I], and Drinking Water Research [DW; LTG II] whereas the
fourth has a major focus on the incorporation of biologically based MOA information into
quantitative risk assessment (the Human Health Research Strategy and the Human Health
MYP; LTG II). Consideration also is being given to how these data and analyses can be
incorporated into next-generation risk assessments, in association with ORD's Human Health
Risk Assessment program and NCEA. NCCT management meets at least quarterly with the
NPDs for these MYPs, in part to continue dialogue on sharing and coordination of resources
among programs and ORD laboratories and centers beyond NCCT. Besides financial
resources, NCCT is cultivating cross-ORD collaborations through shared postdoctoral fellows
and students through NCCT rotational fellowships for ORD scientists. To date, scientists from
NHEERL, NERL, and NCEA have participated in details of 4 months or longer in NCCT to work
on collaborative research projects.
ORD is in the midst of a transformation to a system of integrated, multidisciplinary (IMD)
research projects. The CTRP is actively engaged in planning and discussions on one such IMD
proposal entitled "Decision Support Tools for Preventing, Reducing, and Managing
Chemical Risks." This IMD project will address the tens of thousands of chemicals and millions
of products that current regulatory decision tools do not have the ability to assess, in terms of
impact on life-stage vulnerability, genetic susceptibility, disproportionate exposures, and
cumulative risk. The project will incorporate some of the predictive, high-throughput tools for
exposure and hazards being developed as part of the CTRP; scale them up; and, with attention
to critical life stage impacts, create prioritization algorithms and next-generation risk
assessments that highlight viable management options for prevention, mitigation, and risk
reduction.
26
-------
EPA CompTox Research Program FY2009 to 2012
Final
3. Tox21: A Federal Partnership Transforming Toxicology
The NRC report on Toxicity Testing in the 21st Century has significant implications for human
health risk assessment, and, to accelerate progress in this area, two NIH institutes and EPA
have entered into a formal collaboration known as Tox21 to identify mechanisms of chemically
induced biological activity, prioritize chemicals for more extensive toxicological evaluation, and
develop more predictive models of in vivo biological response. Consistent with the vision
outlined by Krewski et al. in the NRC report, success in achieving these goals is expected to
result in methods for toxicity testing that are more scientific and cost effective, as well as models
for risk assessment that are more mechanistically based. As a consequence, a reduction or
replacement of animals in regulatory testing is anticipated to occur in parallel with an increased
ability to evaluate the large numbers of chemicals that currently lack adequate toxicological
evaluation. Ultimately, Tox21 is expected to deliver biological activity profiles that are predictive
of in vivo toxicities for the thousands of understudied substances of concern to regulatory
authorities in the United States, as well as in many other countries.
The Tox21 collaboration is being coordinated through a 5-year MOU, which leverages the
strengths of each organization. The MOU builds on the experimental toxicology expertise at the
NTP, headquartered at the NIH/NIEHS; HTS technology of the NIH/NCGC, managed by the
NHGRI; and the computational toxicology capabilities of EPA's NCCT. Each party brings
complementary expertise to bear on the application of novel methodologies to evaluate large
numbers of chemicals for their potential to interact with the myriad of biological processes
relevant to toxicity. A central aspect of Tox21 is the unique capabilities of the NCGCs high-
speed, automated screening robots to simultaneously test thousands of potentially toxic
compounds in biochemical and cell-based HTS assays, and an ability to target this resource
toward environmental health issues. As mentioned by Krewski et al., EPA's ToxCast™ program
is an integral and critical component for achieving the Tox21 goals laid out in the MOU.
To support the goals of Tox21, four focus groups (1) Chemical Selection, (2) Biological
Pathways/Assays, (3) Informatics, and (4) Targeted Testing have been established; these focus
groups represent the different components of the NRC vision described by Krewski et al. The
Chemical Selection group is coordinating the selection of chemicals for the Tox21 compound
library to test at NCGC. A chemical library of nearly 2,400 chemicals selected by NTP and EPA
is already under study at NCGC, and results from several dozen HTS assays are already
available. In the near term, this library will be expanded to approximately 8,400 compounds,
with an additional 1,400 compounds selected by NTP; 2,800 compounds selected by EPA and
provided by the CTRP; and 2,800 clinically approved drugs selected by NCGC. Compound
selection currently is based largely on the compound having a defined chemical structure and
known purity, on the extent of its solubility and stability in dimethyl sulfoxide; (the preferred
solvent for HTS assays conducted at the NCGC), and on the compound having low volatility.
Implementing quality control procedures for ensuring identity, purity, and stability of all
compounds in the library is an important responsibility of this group. A subset of the Tox21
chemical library will be included in Phase II of the ToxCast™ program, which will examine a
broader suite of assays to evaluate the predictive power of bioactivity signatures derived in
Phase I.
The Biological Pathways and Assays group is identifying critical cellular toxicity pathways for
interrogation using biochemical- and cell-based high-throughput screens and prioritizing HTS
assays for use at NCGC. Assays already performed at NCGC include those to assess (1)
cytotoxicity and activation of caspases in a number of human and rodent cell types, (2) up-
regulation of tumor suppressor p53, (3) agonist/antagonist activity for a number of nuclear
receptors, and (4) differential cytotoxicity in several cell lines associated with an inability to
27
-------
EPA CompTox Research Program FY2009 to 2012
Final
repair various classes of DNA damage. Other assays under consideration include those for a
variety of physiologically important molecular pathways (e.g., cellular stress responses), as well,
as methods for integrating human and rodent hepatic metabolic activation into reporter gene
assays. Based on the results obtained, this group will construct test batteries useful for
identifying hazard for humans and for prioritizing chemicals for further, more in-depth evaluation.
As Tox21 progresses, it will offer an excellent opportunity to incorporate assays specifically
relevant to the assessment of chemical hazard to wildlife. For example, as assays become
available for the reactivity of nuclear hormone receptors (e.g., estrogen, androgen) from multiple
species, we will have the ability to directly compare their responsiveness and test whether a
particular species might be more sensitive to perturbations than others.
The Informatics Group is developing databases to store all Tox21-related data and evaluating
the results obtained from testing conducted at NCGC and via ToxCast™ for predictive toxicity
patterns. To encourage independent evaluations and analyses of the Tox21 test results, all
data, including the comparative animal and human data where available, will be made
accessible publicly via various databases, including EPA's Aggregated Computational
Toxicology Resource (ACToR), NIEHS' Chemical Effects in Biological Systems (CEBS), and the
National Center for Biotechnology Information's PubChem.
As HTS data on compounds with inadequate testing for toxicity become available via Tox21,
there will be a need to test selected compounds in more comprehensive assays. The Targeted
Testing group is developing strategies and capabilities for this purpose using assays that
involve higher order testing systems (e.g., roundworms (Caenorhabditis elegans), zebrafish
embryos, rodents).
In addition to the testing activities, the MOU promotes coordination and sponsorship of
workshops, symposia, and seminars to educate the various stakeholder groups, including
regulatory scientists and the public, with regard to Tox21-related activities.
4. Communicating Computational Toxicology
As the CTRP has matured, communication of progress has developed beyond the publication of
peer-reviewed papers to include implementation of software and databases, Web sites and
other applications. Given the importance of communicating and disseminating the products of
the CTRP, recruitment of a public affairs/communications specialist for NCCT is currently
underway to provide even greater results in this area.
a. EPA Program Office Training and Implementation of Computational Tools
The NCCT and CTRP partners from across ORD have given numerous seminars, held multiple
1- and 2 day workshops, and provided specific training and installation of computational tools for
EPA program offices. These ad hoc approaches are now being formalized into an online menu
of lectures and tutorials on a broad range of computational toxicology topics. Senior scientists
from NCCT, NHEERL, and NERL are contributing to this resource, and a FY2009 NCCT
recruitment of a communication specialist will accelerate development of this effort. In FY2010,
computational toxicology training initially will focus on the tools ready for program office use,
including DSSTox, ACToR, ToxRefDB, and the ToxMiner tool for analyzing ToxCast™ data.
b. Communities of Practice for Chemical Prioritization and Exposure Science
On the scientific level, the NCCT has initiated two CoPs in the areas of Chemical Prioritization
and Exposure Science that are intended to unite practitioners in the designated fields. The
concept of the CoPs was suggested by the BOSC in April 2005 and since has been adopted as
a primary means of communication and integration of activities across ORD, EPA, and outside
entities and stakeholders. These efforts will serve to enhance communication and coordination,
28
-------
EPA CompTox Research Program FY2009 to 2012
Final
develop common standards, promote consistency, evaluate and provide guidance on best
practices, recommend research priorities, and provide training to interested parties.
5. Developing Clients for Virtual Tissues
The virtual tissue projects are developing systems models of the liver and embryo that will make
the data from CTRP databases and chemical prioritization projects more useful in quantitative
risk assessments. Over the course of design and implementation of these systems models,
interim milestones and deliverables will need to be developed and communicated to program
offices. The process of communicating these products and taking feedback from the programs
will serve to identify and establish longer term clients for these ambitious projects that are
utilizing cutting-edge science that is very different from current regulatory practice.
The Virtual Liver (v-Liver™) project actively is engaging program office personnel to address
challenges in MOA elucidation and quantitative dose-response prediction for chronic liver injury.
In FY2009, chemicals that work through nuclear receptor pathways (e.g., CAR, PXR, PPARs)
were chosen for a proof of concept. Information on these chemicals is being used to populate
the v-Liver™ Knowledgebase (v-Liver-KB) and develop a liver simulator model. Program office
staff from OPPTS and NCEA were consulted on the selection of these chemicals from key
classes of pesticides and industrial chemicals. The first deliverable for risk assessors will be v-
Liver-KB, which formally organizes information on normal hepatic functions and their
perturbation by chemical stressors into pathophysiologic states. The v-Liver-KB will be deployed
as an interactive Web-based and desktop tool to intuitively browse and query physiologic
knowledge on chemicals. This then can be used to hypothesize and test putative MOAs and to
link assay results from ToxCast™ and ToxRefDB with other evidence curated from the
literature. This system will provide computable information on key events that transparently
indicate the uncertainties and data gaps and that make inferences on MOA from experimental
data. In addition, we will work closely with risk assessors to customize the system for specific
requirements. Beta versions of the liver simulator will be applied to program office issues
relating to key chemical classes; this will be an intensively collaborative process between CTRP
scientists and OPPTS and NCEA staff. CTRP scientists working on the v-Liver™ project also
will work with OPP staff on retrospective analyses of chronic and cancer in vivo test data, further
introducing OPP to the v-Liver-KB and the simulator as appropriate.
Motivation for the Virtual Embryo (v-Embyo™) is the scientific need to understand mechanisms
of toxicity and predict developmental defects from complex datasets. The research goal is to
simulate embryonic tissues reacting to perturbation across chemical class, system, stage,
genetic makeup, dose, and time. Data input is detailed knowledge of molecular embryology, as
well as high-throughput data from in vitro models on signaling pathways and cellular
phenotypes. Initial efforts have focused on retrospective analyses of regulatory data from
reproductive and developmental toxicity tests in ToxRefDB. This work has been a collaborative
effort with OPP and has fostered working relationships between the CTRP scientists and
program office staff. Next steps are to identify appropriate developmental and reproductive
toxicities that will support development of systems models and are also of regulatory interest to
program offices. This will include the incorporation of high throughput data from
morphogenetically competent in vitro assays, such as the Embryonic Stem Cell Test and
Zebrafish Embryo Test through collaborations with NHEERL and outside partners. Virtual
Embryo's first goals are to create a knowledgebase and simulation engine that enable in silico
reconstruction of key developmental landmarks, which develop by regulating conserved
signaling pathways and cellular processes and that are sensitive to environmental chemicals of
scientific and regulatory interest to EPA program offices.
29
-------
EPA CompTox Research Program FY2009 to 2012
Final
III. CTRP Project Summaries for FY2009 to 2012
A. Intramural Projects Coordinated by NCCT
There are 9 CTRP projects coordinated by investigators in NCCT that collectively comprise the
core of the CTRP moving forward into FY2009 to 2012. Individual project plans for each of
these are appended to this document (see Appendix IV.A). These projects span the source-to-
outcome continuum of toxicology research (Figure 6), providing critical components for next-
generation risk assessments.
Applying Computational Toxicology Along the
Source-to-Outcome Continuum
Source/Stressor Formation
\
Environmental Concentration
ExpoCast
\
v-Tissues
Effect/Outcome
T oxRefDB
Biological Event
ToxCast
J
External Dose Target Dose
Reverse
Toxicokinetics
Figure 6
1. ACToR—Aggregated Computational Toxicology Resource
Lead/Principal Investigator: Richard Judson
ACToR is a Web-based informatics platform, organized at the top level by chemical and
chemical structure, which is indexing, collecting, and organizing many types of data on
environmental chemicals (Judson etal., 2008). Environmental chemicals are defined as those
likely to be in the environment, including all chemicals regulated or tracked by EPA, as well as
related chemicals, such as pharmaceuticals, that find their way into water sources. ACToR is
indexing and linking to data from hundreds of sources, including the EPA; FDA; the Center for
Disease Control and Prevention (CDC); NIH; academic groups; other governmental agencies
(State and national); and international organizations, such as the World Health Organization
(WHO). Information being indexed and gathered includes in vivo toxicity, in vitro bioassay data,
use levels, exposure information, chemical structure, regulatory information, and other
descriptive data. Planning for the project began in mid-FY2007; beta versions were available
inside EPA since early FY2008; and a public version became available in December 2008.
30
-------
EPA CompTox Research Program FY2009 to 2012
Final
ACToR consists of a back-end database and a front-end Web interface built on low-cost,
publicly accessible applications and tools. Over the next 3 years, ACToR will expand to include
more publicly available resources and data, including more information extracted from text
reports and tabularized, and more information on chemical use and exposure. The latter effort
will be coordinated with the efforts of ExpoCast™ and NERL to identify, index, and extract data
from exposure-related resources of highest interest and importance to EPA programs. In
planned upgrades to ACToR, the ability of users to perform flexible searches across different
layers of data will be enhanced, and customized data downloads will be implemented. ACToR
will serve as the primary vehicle to aggregate and publicly disseminate all published data
associated with the ToxCast™, ToxRefDB, and Tox21 research projects. Additionally, the
ToxMiner and NCCT Chemical Repository systems are being developed as part of ACToR.
These are data repositories and data analysis engines for the ToxCast™/Tox21 projects.
2. DSSTox—Chemical Information Technologies in Support of Toxicology Modeling
Lead/Principal Investigator: Ann Richard
The DSSTox project has implemented high-quality data review procedures for standardized
chemical structure annotation, created linkages connecting diverse toxicity resources within and
outside of EPA, and published high-quality EPA chemical inventories and toxicity data files
spanning more than 10,000 substances. The broad utility of DSSTox data files for
cheminformatics and modeling applications provides significant opportunities to influence the
course of predictive modeling strategies and to encourage wider engagement of toxicologists in
toxicity data representation. The DSSTox project will continue efforts to expand chemical data
file offerings into less well-represented areas of toxicology (immunotoxicology, toxicogenomics,
etc.) and provide varied representations of summary toxicity endpoints. In addition, this research
will explore new representations of chemical structure in relation to the biology (e.g., analog
measures, chemical features, chemical classes) and new representations of biological end
points in relation to modeling (e.g., quantitative end points in terms of potency, summarized or
grouped effects, qualitative active and inactive classes). These efforts will be designed to
complement and augment projects in NCCT (ToxCast™, ACToR, and Tox21) that are working
to improve capabilities to access, mine, and integrate chemical-biological activity information
from existing and new data, both within and outside EPA, in support of toxicity prediction efforts.
In close coordination with ACToR, which is the primary informatics resource for ToxCast™ and
Tox21 chemical and biological data, DSSTox will provide the initial chemical registration
identifications, structure annotation, and quality review of ToxCast™ and ToxRef inventories, as
well as the expanded Tox21 chemical testing library, helping to ensure quality and consistency
of chemical information across the various NCCT programs.
3. ToxRefDB—Toxicity Reference Database
Lead/Principal Investigator: Matthew Martin
Thirty years of registration toxicity data and open literature studies have been archived as
hardcopy and scanned documents by the EPA and others. A significant portion of these data
now have been processed into standardized and structured toxicity data within EPA's
ToxRefDB, including chronic, cancer, developmental, and reproductive studies from laboratory
animals (Martin et al., 2009a ; Martin et al., 2009b; Knudsen et al., 2009). ToxRefDB is a
collaborative project between NCCT and OPP. These data are now accessible and mineable
within ToxRefDB and are serving as a primary source of validation for EPA's ToxCast™
research program in predictive toxicology. In addition to providing reference toxicity information
to research efforts, ToxRefDB will be mined for information on the role and impact of previous
and current study guidelines on the regulation of environmental chemicals. The initial collection
31
-------
EPA CompTox Research Program FY2009 to 2012
Final
of studies in March 2006 focused primarily on the reviews of registrant-submitted toxicity studies
on pesticide active ingredients. ToxRefDB design, development, and implementation were
completed in mid-FY2006 with ongoing updates to the standardized vocabulary and data entry
tool interface. The entry of more than 2,000 studies spanning the majority of the ToxCast™
Phase I chemical set was completed in late FY2008. The status of these initial datasets is either
published, in press, or submitted and is being made publicly available through the ToxRefDB
Web site. A Web-based query tool for the entire contents of ToxRefDB will become available to
the public in 2009, in conjunction with a quarterly update of the EPA ACToR program.
ToxRefDB will continue to enter available data from chronic, cancer, developmental, and
reproductive studies with a focus on potential ToxCast™ Phase II chemicals. The availability
and entry of toxicity data into ToxRefDB also will guide the selection of ToxCast™ Phase II
chemicals. Over the next year, ToxRefDB also will be expanded to capture developmental
neurotoxicity (DNT) study data and possibly in vivo data submitted to the EDSP. In addition,
ToxRefDB is being used for retrospective analyses by various EPA, OECD, and other groups
working on revisions to animal test guidelines and other projects.
4. ChemModel—Application of Molecular Modeling to Assessing Chemical Toxicity
Lead/Principal Investigator: James Rabinowitz
This project is using modern molecular modeling methods developed for the discovery of novel
pharmaceutical agents to computationally predict toxicant-biomolecular target interactions. A
library of computational models of relevant biomolecular targets is being developed. Molecular
modeling approaches may then be used to interrogate this library for the capacity of specific
environmental molecules to interact with each target. The endocrine system provides a test for
the utility of this approach because many of the pathways for toxicity and the macromolecular
targets in those pathways have been identified. Appropriate experimental crystal structures of
many of the receptor protein targets are available to create the computational library of targets.
The ultimate objective of this research is to develop a library of biomolecular targets for
chemical toxicity and the methods appropriate for their application to predicting the capacity of a
chemical to interact with these targets. This library of targets then may be used in conjunction
with other approaches as part of a chemical prescreen.
5. ToxCast™—Screening and Prioritization of Environmental Chemicals Based on
Bioactivity Profiling and Predictions of Toxicity
Lead/Principal Investigator: Keith Houck
The objective of the ToxCast™ research program is to develop a cost-effective and rapid
approach for screening and prioritizing a large number of chemicals for toxicological testing.
Using data from HTS bioassays developed in the drug discovery field, ToxCast™ is generating
data, constructing databases, and building computational models to forecast the potential
human toxicity of chemicals. HTS bioassays for ToxCast™ also are being provided by NHEERL
partners. These hazard predictions should provide EPA regulatory programs, including OPP,
with science-based information helpful in prioritizing chemicals for more detailed toxicological
evaluations, ultimately leading to reduced animal testing. Furthermore, the toxicity pathways
identified from this dataset and project will be critical to transforming the practice of risk
assessment for environmental chemicals and contaminants (in collaboration with NCEA and
EPA program offices). ToxCast™ is a multiyear effort that is divided into three distinct phases.
• Phase I: 300 Chemicals assayed in more than 600 different HTS bioassays, to create
predictive bioactivity signatures based on the known toxicity of the chemicals
• Phase II: Focused on confirmation and expansion of ToxCast™ predictive signatures,
generating HTS data on 700 additional chemicals
32
-------
EPA CompTox Research Program FY2009 to 2012
Final
• Phase III: ToxCast™ expanded to thousands of environmental chemicals for which little
toxicological information is available
Once ToxCast™ has gone through successful initiation of Phase III, the data and toxicity
predictions will be ready for deployment throughout numerous EPA program offices. NCCT will
work to link these hazard predictions with exposure predictions and create integrated database
analysis tools facilitating customized chemical prioritizations appropriate to specific programs.
Beyond the initial application of ToxCast™ data and tools to prioritizing chemicals for further
screening, testing, and monitoring, secondary applications will include the Virtual Tissues
systems modeling projects and next-generation risk assessments with NCEA, NHEERL, and
EPA program offices.
6. ExpoCast™—Exposure Science for Screening, Prioritization, and Toxicity Testing
Lead/Principal Investigator: Elaine Cohen Hubal
ExpoCast™ will provide an overarching framework for the science required to characterize
biologically relevant exposure as a critical part of the CTRP (Cohen Hubal, 2009). The
ExpoCast™ program will foster novel exposure science research to (1) inform chemical
prioritization, (2) understand system response to chemical perturbations and implications at the
individual and population levels, and (3) link information on potential toxicity of environmental
contaminants to real-world health outcomes. An important early component of ExpoCast™ will
be to consider how best to consolidate and link human exposure data for chemical prioritization
and toxicity testing. ExpoCast™ represents a strong collaboration between NCCT and NERL,
with both parties providing leadership and critical scientific contributions toward this
transformation of exposure science. Initial research will focus on identifying and evaluating
novel approaches for characterizing exposure to prioritize chemicals and developing modeling
approaches for considering exposure potential based on chemical properties, sources (e.g.,
consumer products), uses, lifecycle, and individual and population vulnerability. Later
applications of ExpoCast™ data and tools will include next-generation risk assessments with
NCEA, NERL, and EPA program offices.
7. v-Embryo™—Virtual Embryo
Lead/Principal Investigator: Thomas Knudsen
Motivation of the Virtual Embryo Project (v-Embyro™) is scientific needs to understand
mechanisms of toxicity and predict developmental defects from complex datasets. EPA must
evaluate environmental chemicals for potential effects on development. Part of this challenge is
to understand mechanisms by which chemicals disrupt prenatal development. Unfortunately, the
mechanisms of prenatal developmental toxicity are not understood in sufficient depth or detail
for risk assessment purposes. Because embryonic tissues are regulated simultaneously by
pathways that control genetic patterning, molecular clocks, morphogenetic tissue
rearrangements, and cellular differentiation, there is a need for computational (in silico) models
to address this complexity. EPA's v-Embyro™ will comprise a framework to merge data and
knowledge about developmental processes, leading to cell-based computational models that
can be used to analyze mechanisms in developmental toxicity. This is a collaborative project
among NCCT, NHEERL, and NCEA. Data input is detailed knowledge of molecular embryology,
high-throughput data from in vitro models, signaling pathways, and cellular phenotypes. Output
models aim for the modular reconstruction of a developing embryo from cell-based models of
morphogenesis and differentiation.
33
-------
EPA CompTox Research Program FY2009 to 2012
Final
8. v-Liver™ —The Virtual Liver Project
Lead/Principal Investigator: Imran Shah
The Virtual Liver Project (v-Liver™) computational paradigm represents tissues as cellular
systems in which discrete individual cell-level responses give rise to complex physiologic
outcomes. In this model, cell-level responses are governed by a self-regulating network of
normal molecular processes, and adverse histopathologic effects arise because of chronic
stimulation by environmental chemicals. The v-Liver™ proof of concept is being developed by
(1) focusing on environmental chemicals responsible for hepatocarcinogenesis in rodent
studies, (2) organizing MOA knowledge on the relevant molecular and cellular processes
perturbed by these chemicals, (3) developing a tissue simulation platform to investigate the
uncertainties in MOA and neoplastic lesion formation, and (4) evaluating in vitro assays to
predict lesion development across chemicals and doses. The Virtual Liver Project is
collaboration between NCCT and NHEERL.
9. Uncertainty—Analysis in Toxicological Modeling
Lead/Principal Investigator: R. Wood row Setzer
The goals of this project are to develop standardized and more efficient computational
approaches for parameter estimation and model selection; standardize approaches for model
evaluation for PBPK and other dynamic models; and develop methods for constructing priors
(probabilistic summaries of current knowledge) for model parameters, based on existing
computational methods and datasets. The initial motivation for this work was the need to
standardize and make more sophisticated parameter estimation and model evaluation for PBPK
models being used by OPP, and that emphasis will continue in the early phase of this project.
However, all models relevant to toxicological risk assessment have similar requirements, and
this project will coordinate closely with the Virtual Tissues and ToxCast™ projects. In particular,
the project will collaborate with ToxCast™ in developing approaches for quantifying uncertainty
in ToxCast™ predictions and prioritizations.
B. Intramural Projects Coordinated by NERL and NHEERL
Several key components and research projects of the CTRP are coordinated by NERL,
NHEERL or both in conjunction with NCCT and other collaborators.
1. NERL
NERL is conducting human exposure research for screening, prioritization, and toxicity testing in
collaboration with and complementary to the ExpoCast™ project. Exposure science is crucial for
addressing many of our important and complex environmental health issues and is essential for
toxicity testing to be valuable in public health protection. There is a clear need for a collaborative
effort across the exposure and risk assessment community to ensure the required exposure
science, data, and tools are ready to address immediate needs resulting from application of
high-throughput in vitro technologies for toxicity testing. A coherent program is required to
formulate significant exposure questions posed by these novel in vitro toxicity data; develop
creative approaches for applying existing exposure information and tools to address these
questions; and, finally, identify key exposure research needs to interpret the toxicity data for risk
assessment. The authors of the National Academies report (NRC, 2007) emphasize that
population-based data and human exposure information are required at each step of their vision
for toxicity testing and risk assessment. The collaborating NERL and NCCT scientists have
34
-------
EPA CompTox Research Program FY2009 to 2012
Final
identified and will be conducting research in the following priority exposure research areas to
support chemical screening, prioritization, and toxicity testing (Sheldon and Cohen Hubal,
2009}: (1) accessible and linkable exposure databases, (2) exposure screening tools for
accelerated chemical prioritization, (3) computational tools for dose reconstruction and source-
to-outcome analyses, (4) tools for understanding the fundamental processes and factors
influencing human exposures, and (5) efficient monitoring methods to measure and interpret
biologically relevant exposure metrics. Susceptibility, vulnerability, and life-stage aspects are
integral to each of these.
The NERL-directed aspects of the ExpoCast™ program will include research required to
understand fundamental processes and factors influencing human exposures, as well as
development of the tools required to facilitate efficient exposure assessment. This research will
be implemented in the following broad areas.
Prioritization and screening. Two related research activities will be implemented: (1) developing
high-quality, high-quantity exposure databases aligned with the NCCT databases and (2)
developing and evaluating screening models for risk assessment. NERL will inventory, compile,
and organize available environmental and exposure data into readily accessible databases.
These databases will be organized efficiently so that they are aligned and integrated with other
NCCT databases, allowing the users to link source, environmental, exposure, and effects data
for chemicals and degradates/metabolites. Research will be conducted to develop the next
generation of predictive environmental fate and transport, exposure, and dose-screening
models that can be linked with the corresponding NCCT, NHEERL, and other Agency toxicity
screening-level models. Available screening models will be inventoried and assessed. A
workshop of international experts will be scheduled for evaluating the available models currently
being used by various international organizations. New and refined models will be developed,
based on the evaluation results, and linked with appropriate Agency toxicity models for future
risk assessments.
Linked exposure-dose models. Research will be conducted to efficiently link NERL's
environmental, exposure and dose models and databases for supporting program office specific
exposure and risk assessments. Emphasis will be on developing tools and approaches to
facilitate rapid assessment.
Biomarkers. Collaborative research with scientists from ORD, CDC, and academia will be
implemented to design studies for developing and evaluating tools to interpret the results of
exposure biomarker studies and link these results to indicators of first biological response. The
Metabolic Simulator and Metapath research tools will be upgraded, and their utility for risk
assessments will be evaluated using industry-provided data. Collaborative observational studies
will be conducted with CDC, NHEERL, and others to develop and refine models for relating
measured exposure biomarker results with environmental exposures (both forward and reverse
dosimetry). Research will be conducted with NCCT and NHEERL to understand how "omics"-
based exposure biomarker data, combined with chemical biomarker results, can be used to link
exposure with indicators of effects.
Observational studies. Collaborative research with The National Children's Study and STAR
awardees will be conducted to identify and characterize the key factors influencing children's
exposures to pesticides and other chemicals. Research activities also will be implemented to
characterize real-world exposures to multiple chemicals (mixtures) in targeted communities and
vulnerable populations (including children). Tools will be developed for predicting high
35
-------
EPA CompTox Research Program FY2009 to 2012
Final
exposures, understanding the factors contributing to these exposures, and providing input for
the development and evaluation of risk reduction strategies.
2. NHEERL
Research within NHEERL is both parallel to and integrated with the CTRP. Although some of
this research has been ongoing for a number of years, a new effort to expand this program
currently is underway. The overall goal is the prediction of chemical toxicity to humans and
wildlife based on understanding fundamental biology and its perturbation by toxicants. The
approach is based on the elucidation of key events that link initiating events to adverse
outcomes and leverages expertise in human studies, whole animal toxicology in a wide range of
species, and cellular and molecular biology to identify toxicity pathways associated with adverse
health and ecological outcomes. The five focus areas for NHEERL research and the integration
of these efforts into the CTRP are described below.
(1) Linkage of environmental exposure to perturbation of toxicity pathways. For the appropriate
application of high-throughput assays based on toxicity pathways to chemical screening
(hazard identification) and for risk assessment, the environmental exposure levels and their
relation to the exposure at the cellular level must be known. Pharmacokinetic studies and
modes will be used to determine the relationship between external exposures and tissue
dose in vivo, whereas cellular and molecular biology studies of in vivo effects and
subsequent development of toxicity pathway assays will broaden the scope of screening
assays.
(2) Linkage of toxicity pathway perturbations to adverse outcomes. Research here is focused
on identifying toxicity pathways, key events, and MOAs as they relate to adverse outcomes
and disease. MOA is defined as a sequence of key events and processes, beginning with
interaction of an agent with a cell, proceeding through operational and anatomical changes,
and resulting in an adverse health effect. Global biology measures ("omics") will be used to
discover new toxicity pathways that then will be translated into medium-and high-throughput
in vitro and non-mammalian screening assays. Results from these new technologies will be
compared with predictions from in vivo experiments. In cases where molecular targets are
established for chemical classes, quantitative structure-activity relationship (QSAR) models
and read-across methods will be developed to predict the toxicological potential of untested
chemicals. These efforts will be coordinated with activities in OPPTS, OECD, and the
European Union to ensure the efforts meet Agency and international needs for regulatory
purposes.
(3) Development of toxicity pathway assays. NHEERL is developing assays for
neurodevelopmental, immunotoxicity, and cellular stress responses. As these assays are
established, these will be used to expand the breadth of assays in the ToxCast testing
program. To date, the ToxCast™ Phase I chemicals have been tested in several
neurodevelopmental and cellular stress assays. Through participation in the Tox21 MOU,
NHEERL has contributed additional toxicity pathway assays related to the assessment of
stress responses in cellular systems. In addition, NHEERL will be performing secondary
screening and targeted testing to explore insights and hypotheses generated from the
ToxCast and Tox21 HTS efforts. These follow-up studies are designed to evaluate findings
from the screening and provide quantitative in vivo relationships.
(4) Quantitative models for risk assessment. As quantitative relationships are established
between assay conditions and environmental exposures for humans and wildlife, a transition
to toxicity-pathway-based risk assessment becomes technically feasible. Data to define
these quantitative relationships will be generated in collaboration with NCEA and other ORD
partners to ensure the suitability for use in quantitative modeling. Research integrated with
36
-------
EPA CompTox Research Program FY2009 to 2012
Final
NERL modeling efforts will make PK models compatible with exposure models. Modeling of
MOA is being conducted in collaboration with NCCT through modeling of virtual
tissues/systems/organisms, as well as less detailed BBDR modeling projects. Through the
integrated nature of this research, it is anticipated that in vivo models will be generated for
comparison with in vitro toxicity pathway screening results. In addition, models will be
developed that can take in vitro results directly as input and make quantitative in vivo
predictions for target organisms. The v-Liver™ and v-Embryo™ projects are two NCCT-
NHEERL collaborations currently underway in this area. An additional virtual
cardiopulmonary project currently is being developed based on extensive NHEERL research
in this area and the CTRP "new start" project is looking at the mechanistic indicators of
childhood asthma (MICA).
(5) Special considerations for ecology research. The actions presented thus far are applicable
to both health and ecological problems. Ecology has some specific issues, however, that
must be addressed in parallel with the efforts described above. First, although human health
risk assessment can be conducted on the basis of a susceptible subpopulation's risk of an
adverse outcome, most ecological concerns relate to the effects on population viability.
There has been demonstrated success within NHEERL of linking MOA models to population
models that will be extended to toxicity pathway-based models to be used for ecological risk
assessment as relevant. Second, the NRC report highlights the many problems in
extrapolating among species and recommends that species extrapolation be avoided by
focusing on human cells for toxicity pathway assays. This is not possible for ecological risk
assessment, as the number of relevant species is much too large for direct testing in each
one. Therefore, the identification of appropriate sentinel species and development of toxicity
pathway assays in these species will be coupled with the development of methods for
species extrapolation.
C. Extramural STAR Grantee Projects
Implicit in many of the research projects contained within the CTRP, bioinformatics is one area
of research that needs more focus and attention. This rapidly emerging technology is crucial to
the computational toxicology program and there remains a large gap in ORD relative to the
ability to analyze the high volumes of molecular data and to predict potential toxicity; MOAs;
and, ultimately, risk. To help bridge this gap, NCER has supported the establishment of two
STAR Environmental Bioinformatics Research Centers. The Research Center for Environmental
Bioinformatics and Computational Toxicology at the University of Medicine and Dentistry of New
Jersey (UMDNJ), Piscataway, NJ, and The Carolina Environmental Bioinformatics Research
Center at the University of North Carolina, Chapel Hill, are operating as cooperative agreements
and helping to facilitate the application of bioinformatics tools and approaches to environmental
health issues supported by the CTRP.
In the next year, UMDNJ researchers will begin the design of new ebTrack interfaces to open
source databases and to various "external" and center-developed modeling tools for facilitating
wider deployment and applicability of the ebTrack/ArrayTrack system for integrative analyses of
various types of genomic, proteomic, and metabonomic data. Additionally, plans are underway
to refine the environmental bioinformatics knowledgebase (ebKB) and to make a public beta
version of ebKB available; implement a modular "Virtual Liver" with alternative levels of detail in
describing physical structure of the liver with respect to toxicokinetic and toxicodynamic
processes with case studies focusing on environmentally relevant chemicals; and refine the
framework for Dose-Response Information Analysis (DORIAN) modules representing different
37
-------
EPA CompTox Research Program FY2009 to 2012
Final
scales of biological complexity, ranging from molecule-molecule interactions to biochemical
networks to virtual organs and systems.
For the Carolina Bioinformatics Center, goals for the next year include (1) continuing progress in
dose-response pathway modeling and analysis of ToxCast™ Phase I data; (2) continuing QSAR
modeling of multiple animal toxicity end points and (3) the development of QSAR and other
statistical models to use in vitro biological data to predict in vivo toxicity end points. Other plans
include the development of specific data-mining algorithms for genomic databases, and
extending computational work on fast approaches for genome-wide expression QTL analysis to
human haplotypes.
The third STAR center is the Carolina Center for Computational Toxicology at the University of
North Carolina, Chapel Hill. This center is developing fine-scale predictive simulations of the
protein-protein/-chemical interactions in nuclear receptor networks, mapping chemical-perturbed
networks and devising modeling tools that can predict the pathobiology of compounds based on
a limited set of biological data, building tools that will enable toxicologists, to understand the role
of genetic diversity between individuals in responses to toxicants; and creating unbiased
discovery-driven prediction of adverse chronic in vivo outcomes based on statistical modeling of
chemical structures and HTS.
Another high-priority for EPA is to understand the molecular and cellular processes that, when
perturbed, result in developmental toxicity. In response to this need, NCER has funded the
Texas-Indiana Virtual STAR Center; Data-Generating in vitro and in silico Models of
Developmental Toxicity in Embryonic Stem Cells and Zebrafish at the University of Houston -
University Park, Indiana University - Bloomington, and Texas A & M University Institute for
Genomic Medicine. As chemical production increases worldwide, there is a concordant
increased need for determining the hazard and risk to human health at realistic exposure levels.
The main objective of the proposed multidisciplinary Texas-Indiana Virtual STAR Center; Data
Generating in vitro and in silico Models of Developmental Toxicity in Embryonic Stem Cells and
Zebrafish (TIVS) is to contribute to a more reliable chemical risk assessment through the
development of high-throughput in vitro and in silico screening models of developmental toxicity.
Specifically, the TIVS Center aims to generate in vitro models of murine embryonic stem cells
and zebrafish for developmental toxicity. The data produced from these models will be f
exploited urther to produce predictive in silico models for developmental toxicity on processes
that are also relevant for human embryonic development.
D. Summary Integration of the CTRP Projects for FY2009 to 2012
The CTRP spans several ORD laboratories and centers, as well as the extramural STAR grants
program. Collectively, these various components of the CTRP are developing new methods and
tools that will enhance our ability to predict adverse effects and understand the mechanisms
through which chemicals induce harm. Advances from the CTRP will give EPA the ability to
screen and assess a larger number of chemicals than traditional methods enable. In addition,
EPA is collaborating with other governmental and private organizations to leverage resources
and access complementary expertise to accelerate progress in high-priority research areas.
Throughout the various components of the CTRP, focus is maintained on addressing a number
of key science questions.
What are the key linkages in the continuum between the source of a chemical in the
environment and its adverse outcomes?
How can we develop predictive models for screening and testing?
38
-------
EPA CompTox Research Program FY2009 to 2012
Final
How can we improve quantitative risk assessment and reduce uncertainty by using advanced
computational techniques?
How can we enhance dose-response modeling, especially in low-dose ranges, to include
knowledge of molecular events?
To address these questions, ORD's CTRP will continue to provide informatics, chemical
prioritization, and systems modeling solutions for EPA. ACToR, DSSTox, ToxRefDB,
ToxCast™, and ExpoCast™, v-Liver™, and v-Embryo™ are all excellent examples of
extensive, multidisciplinary and integrated projects that bring together the talents of ORD, along
with extramural scientists from EPA-funded STAR centers to provide the high-throughput
decision support tools for screening and assessing chemical exposure, hazard, and risk.
39
-------
EPA CompTox Research Program FY2009 to 2012
Final
IV. APPENDIXES
A. Intramural CTRP Projects
1. Project Plans
a. ACToR—Aggregated Computational Toxicology Resource
Lead/Principal Investigator: Richard Judson
Version: March 4, 2009
Research Issue/Relevance: EPA faces a significant issue in that there are many chemicals in
wide-spread use, and which the Agency regulates, for which there is little or no toxicology
information. The EPA ToxCast™ program is one effort that is addressing this problem by
screening many of these chemicals using high-throughput techniques and helping prioritize
which ones are candidates for more detailed testing. To be effective, ToxCast™ needs to know
what chemicals are in need of screening and needs to know what is already known (or not)
about these chemicals. ACToR is providing this information. In addition, the ToxCast™,
ToxRefDB, and Tox21 projects each have need for unified, sophisticated informatics support
and management of the large amount of chemical and biological information that is central to
these projects. Additional capabilities tied to this informatics resource will be needed to address
the data analysis and toxicity prediction challenges of the ToxCast™/Tox21 projects.
A related issue is that other EPA programs, as well as external stakeholders, need easy access
to information on environmental chemicals both within and beyond the set of interest to
ToxCast™. Currently, toxicity and exposure data associated with environmental chemicals does
not adhere to standardized representations and are dispersed widely across many databases
and Internet resources, many of which are difficult to access or search. The ACToR project is
addressing this challenge by creating a central, standardized, publicly accessible chemical-
informatics platform to enable searching and cross-referencing of chemical-associated toxicity
information to aid prioritization and hazard identification of environmental chemicals.
Purpose/Objective/Impact: ACToR aims to provide a unified, centralized resource of data on
environmental chemicals, including toxicology, in vitro assay data, and chemical structure
information. By gathering information on the type and location of toxicity or exposure data
associated with environmental chemicals into a single, searchable, publicly accessible Web-site,
ACToR is providing the basis for chemical selection and screening within NCCT projects, such
as ToxCast™ and Tox21. ACToR also is coordinating with the DSSTox project to incorporate
quality chemical review and structure-annotation for the chemical datasets of highest interest to
the various NCCT projects. In addition to its use in supporting various NCCT and EPA projects,
ACToR is a publicly available EPA resource that enables other government agencies, industry,
and academic researchers to quickly search and collate toxicity-related information on
chemicals of interest. As such, it will promote and encourage other entities to adopt standards
for chemical representation and broadly survey chemical information pertaining to toxicology
resources on the Internet.
Synopsis: ACToR is a Web-based informatics platform, organized at the top level by chemical
and chemical structure that is indexing, collecting, and organizing many types of data on
environmental chemicals. Environmental chemicals are defined as those likely to be in the
40
-------
EPA CompTox Research Program FY2009 to 2012
Final
environment, including all chemicals regulated or tracked by EPA, as well as related chemicals,
such as pharmaceuticals that find their way into water sources. ACToR is indexing and linking to
data from hundreds of sources, including the EPA, FDA, CDC, NIH, academic groups, other
governmental agencies (State and national) and international organizations, such as the WHO.
Information being indexed and gathered includes in vivo toxicity, in vitro bioassay data, use
levels, exposure information, chemical structure, regulatory information, and other descriptive
data. Planning for the project began in mid-FY2007; beta versions were available inside EPA
beginning in early FY2008, and a public version became available in December 2008. ACToR
consists of a back-end database and a front-end Web interface built on low-cost, publicly
accessible applications and tools. Over the next 3 years, ACToR will expand to include more
publicly available resources and data, including more information extracted from text reports
and tabularized, and more information on chemical use and exposure. The latter effort will be
coordinated with the efforts of ExpoCast™ and NERL to identify, index, and extract data from
exposure-related resources of highest interest and importance to EPA programs. In planned
upgrades to ACToR, the ability of users to perform flexible searches across different layers of
data will be enhanced, and customized data downloads will be implemented. ACToR will serve
as the primary vehicle to aggregate and publicly disseminate all published data associated with
the ToxCast™, ToxRef, and Tox21 research projects. Additionally, the ToxMiner and NCCT
Chemical Repository systems are being developed as part of ACToR. These are data
repositories and data analysis engines for the ToxCast/Tox21 projects.
Partnerships/Collaborations (Internal & External):
(1) EPA ToxCast™—program—provide data for use in selecting chemicals and providing
toxicology data for validation; provide route for publication of data
(2) Tox21 partnership—provide data for use in selecting chemicals and providing toxicology
data for validation; provide route for publication of data
(3) DSSTox coordination—align methods for registering high-interest chemical inventories
(ToxCast™, ToxRefDB, Tox21, DSSTox published data files), utilizing DSSTox chemical
information quality review and structure-annotation within ACToR
(4) EPA centers and offices (OPPT/OPP/NCEA/OW)—provide data on chemicals of interest
Milestones/Products:
FY09
(1) Initial public deployment
(2) Significant version 2, including refined chemical structure information
(3) Develop workflow for tabularization of data buried in text reports
(4) Integrate all ToxCast™ and ToxRefDB data
(5) Quarterly releases with new data
FY10
(1) Quarterly releases with new data
(2) Implementation of a process to gather tabular data on priority chemicals from text reports
(3) Survey sources of chemical use and exposure data and import any remaining sources
(4) Develop flexible query interface and data download process
(5) Develop process to extract data from open literature
FY11
(1) Quarterly releases with new data
FY12
(1) Quarterly releases with new data
Keywords: computational toxicology, ACToR, ToxCast™, DSSTox, database
41
-------
EPA CompTox Research Program FY2009 to 2012
Final
QA Project Plan: Category II. ACToR development is guided by a series of standard operating
procedures. These govern all aspects of the project, including data acquisition, formatting,
quality assurance, database filling and maintenance, and system administration. All quality
assurance (QA) plans are archived in the EPA internal QA system.
b. DSSTox—Chemical Information Technologies in Support of Toxicology Modeling
Lead/Principal Investigator: Ann M. Richard
Research Issue/Relevance: A central regulatory mandate of EPA is to assess the potential
health and environmental risks of large numbers of chemicals released into the environment,
often in the absence of relevant test data. Significant advances in toxicity prediction capabilities
are predicated on the ability to store, mine, and analyze information on many levels in relation to
chemicals and their effects on biological systems. Standardized high-quality chemical structure
annotation and searchability of toxicity-related information across the Internet and within EPA
programs is a crucial requirement for creating effective data linkages and gathering relevant
data. Equally important is the need to incorporate meaningful chemical structure and property
representations, based on principles of organic chemistry and biologically informed measures of
chemical similarity, into toxicity modeling efforts. Finally, successful modeling efforts will depend
on suitable representations of biological activity, both HTS and in vivo, in relation to chemical
structure. NCCT's ToxCast™ and Tox21 projects are employing HTS tests to probe biochemical
target interactions, chemical pathways, and cellular responses potentially relevant to toxicity for
thousands of chemicals of high potential exposure and environmental interest. The goal is to
use these data, in conjunction with legacy data and chemical structure considerations, to infer
meaningful patterns and to develop models to predict a range of in vivo bioassay responses.
Biologically informed toxicity prediction models that incorporate chemical structure-activity
considerations are likely to provide the best means for prioritizing large lists of chemicals for
potential hazard—a pressing need for many EPA programs—and charting a path forward for a
more efficient and cost-effective screening and testing paradigm.
Purpose/Objective/Impact: The current research will use the DSSTox project framework to
incorporate strict quality standards for chemical information across NCCT projects (ToxCast™,
ACToR, and Tox21) and expand comparability and linkages of summary toxicity data in the
context of standardized cheminformatics environments. This project also will use these data
foundations to promote and explore new ways to associate chemical structure with biological
activity, extending traditional SAR toward new paradigms for biologically informed structure-
based toxicity prediction. These efforts have the potential to impact a wide variety of EPA
program offices that rely heavily on chemical information resources and have a need for
structure-based data exploration, analog searching, and improved toxicity prediction models
when limited test data are available. These include programs within OPPTS e.g., Green
Chemistry, PreManufacture-Notification Program (PMN), OPP, HPV Testing Program, as well
as EPA's IRIS Program, Office of Water, and Office of Environmental Information. New
information technologies that incorporate strict quality standards and more flexible and diverse
means for assessing biological and chemical similarity also will improve the identification of
toxicologically relevant analogs by enhancing the ability to explore data and quantify
associations across diverse chemical and biological data domains.
Synopsis: The DSSTox project has implemented high-quality data review procedures for
standardized chemical structure annotation, created linkages connecting diverse toxicity
resources within and outside of EPA, and published high-quality EPA chemical inventories and
42
-------
EPA CompTox Research Program FY2009 to 2012
Final
toxicity data files spanning more than 10,000 substances. The broad utility of DSSTox data files
for cheminformatics and modeling applications provides significant opportunities to influence the
course of predictive modeling strategies and to encourage wider engagement of toxicologists in
toxicity data representation. The DSSTox project will continue efforts to expand chemical data
file offerings into less well-represented areas of toxicology (immunotoxicology, toxicogenomics,
etc.), and provide varied representations of summary toxicity end points. In addition, this
research will explore new representations of chemical structure in relation to the biology (e.g.,
analog measures, chemical features, chemical classes) and new representations of biological
end points in relation to modeling (e.g., quantitative end points in terms of potency, summarized
or grouped effects, qualitative active and inactive classes). These efforts will be designed to
complement and augment projects in NCCT (ToxCast™, ACToR, and Tox21) that are working
to improve capabilities to access, mine, and integrate chemical-biological activity information
from existing and new data, both within and outside EPA, in support of toxicity prediction efforts.
In close coordination with ACToR, which is the primary informatics resource for ToxCast™ and
Tox21 chemical and biological data, DSSTox will provide the initial chemical registration IDs,
structure annotation, and quality review of ToxCast™ and ToxRefDB inventories, as well as the
expanded Tox21 chemical testing library, helping to ensure quality and consistency of chemical
information across the various NCCT programs.
Partnerships/Collaborations (Internal & External): The DSSTox project is being coordinated
and linked with a number of public efforts (ILSI, ToxML, LHASA UK, PubChem, ChemSpider)
and government research laboratories (NIEHS, NTP, FDA) that are promoting controlled toxicity
vocabularies, adopting data standards, and migrating diverse toxicity data into the public
domain. DSSTox also is aligned with major NCCT projects (ToxCast™, ToxRefDB, ACToR,
Tox21), providing key quality review procedures and cheminformatics support, expanding
DSSTox data file publications of toxicological data in support of predictive modeling, and
enhancing linkages to public resources such as PubChem, for disseminating bioassay results to
the broader modeling community. In coordination with ACToR, partnerships and collaborations
with scientists across EPA (NHEERL, OPPT, OPP, and NERL) are being forged to improve
cheminformatic capabilities across the Agency from a unified chemical structure perspective,
most recently extending into exposure data arenas (ExpoCast™). Research collaborations are
ongoing with SAR modelers at UNC (A. Tropsha, H. Zhu) and, in the data mining and SAR
community (C. Yang, R. Benigni, E. Benfenati), to improve methods to incorporate biological
considerations into SAR models. OECD is using DSSTox as a source of high-quality, quality-
controlled chemical structures and activities in their QSAR Toolbox. Finally, we are pursuing
closer collaborations with toxicogenomics resources, such as the National Center for
Biotechnology Information (NCBI; GEO), the European Bioinformatics Institute (EBI: ChEBI and
ArrayExpress), and the NIEHS CEBS toxicogenomics resource.
Milestones/Products:
FY09
(1) Publish paper on ToxCast™ 320 chemical inventory from SAR modeling perspective
(2) Publish papers and coordinate efforts with NCBI (GEO) and EBI (ArrayExpress) to structure-
annotate and provide chemical linkages to microarray data for toxicogenomics
(3) Restart Chemoinformatics CoPs using EPA's Science Portal
(4) Publish DSSTox files for ToxRef and ToxCast™ inventories and selected summary end
points and facilitate publication and linkage to the NLM PubChem project. Compile and
publish public genetic toxicity data and SAR predictions for ToxCast 320
(5) Continue expansion of the DSSTox public toxicity database inventory for use in modeling
(6) Perform primary chemical review and structure annotation of the ToxCast™/Tox21 chemical
testing libraries, coordinating with ACToR and within a central chemical registry
43
-------
EPA CompTox Research Program FY2009 to 2012
Final
FY10
(1) Publish DSSTox files forTox21 inventory and selected summary end points, and facilitate
publication and linkage to the NLM PubChem project
(2) Publish DSSTox files for NTP study areas (Immunotox, Genetox, etc.) to facilitate
incorporation into ACToR and encourage broader SAR examination; explore new
approaches to SAR modeling based on feature categories within existing DSSTox files and
ToxCast™ data
(3) Explore new approaches to SAR modeling based on classifiers and feature categories
(4) Expand CEBS collaboration to incorporate DSSTox GEO and ArrayExpress files; create
chemical linkage to I LSI Developmental Toxicity database and facilitate structure-searching
(5) Advise and assist efforts within ExpoCast™ to identify and chemically annotate important
exposure-related public data resources
FY11
(1) In collaboration with ACToR, establish procedures and protocols for automating chemical
annotation of new experimental data generated by NCCT programs (ToxCast™ and Tox21)
and in collaboration with CEBS or NHEERL
(2) Document and employ PubChem analysis tools in relation to published DSSTox and
ToxCast™ data inventory in PubChem
(3) Collaborate with SAR modeling efforts to predict ToxCast™ end points using in vitro data
(4) Continue expansion of DSSTox public toxicity database inventory for use in modeling with
co-publication and linkage to ACToR and PubChem
FY12
(1) Redesign DSSTox Web site to provide hosting of donated chemical descriptors, properties,
and predictions for high interest inventories
(2) Publish master tables of DSSTox IDs and high-quality structures to serve as public data
registry for toxicology, particularly for EPA, FDA, and NTP datasets
(3) Promote use of chemical registry system from ToxCast™/Tox21, linked to DSSTox content
and integrated into ACToR, more broadly within EPA
(4) Collaborate with SAR modeling efforts to expand modeling to address Tox21 chemicals and
end points
(5) Continue expansion of DSSTox public toxicity database inventory into toxicity and exposure
areas not effectively linked to current databases
Keywords: DSSTox, prediction, cheminformatics, structure-activity, SAR
QA Project Plan: Category II. The DSSTox project, involving the compilation, standardization,
and review of chemical data largely extracted from secondary public data sources has a large
intrinsic QA component. All QA plans are archived in the EPA internal QA system.
Documentation and procedures contributing to overall QA objectives include
• maintenance of a DSSTox Web page describing QA procedures for obtaining and reviewing
chemical information prior to inclusion in the DSSTox Master File
(http://www.epa.gov/ncct/dsstox/ChemicallnfQAProcedures.html);
• file versioning, error tracking within data files, publication of log files with version history, and
full documentation of published DSSTox data files;
• maintenance of the DSSTox Master File tables in ACCESS, cross referencing, and use of
automated scripts to perform field content error checks for new file generation;
• coordinated publication in ACToR and PubChem allows for checks on structure
consistencies;
• review of all newly published DSSTox data files by Source collaborators;
44
-------
EPA CompTox Research Program FY2009 to 2012
Final
• error-reporting system associated with DSSTox published files and structure browser,
c. ToxRefDB—Toxicity Reference Database
Lead/Principal Investigator: Matthew Martin
Version: April 02, 2009
Research Issue/Relevance: As EPA moves toward a new chemical toxicity testing paradigm,
the animal toxicity data the Agency has received will provide context for assessing many of the
new HTS technologies for toxicity testing and screening. ToxRefDB is the relational database
designed and developed to electronically capture all of the relevant information spanning 30
years of health effects data from the Agency and beyond. The EPA ToxCast™ program is one
effort using high-throughput techniques to prioritize chemicals for further testing. To be effective,
ToxCast™ needs reference toxicity information to provide the interpretive context for the large
amounts of screening data. ToxRefDB is providing the reference in vivo toxicity data for
programs such as ToxCast™ in a searchable and computable format.
Additionally, the utility of ToxRefDB is broader than use as a validation dataset for ToxCast™.
Regulatory scientists have begun to assess the role and impact of previous and current
guideline studies and components of those studies in the regulation and assessment of
chemicals. ToxRefDB will be the primary data source for numerous retrospective analyses and
may have a large impact on future revisions to existing guideline studies.
Purpose/Objective/Impact: The ToxRefDB project has provided access to a wealth of in vivo
toxicity data in a structured and searchable format. These data are being released through a
series of manuscripts that currently are submitted for publication or in preparation. In addition,
ToxRefDB data will be publicly available through the ToxRefDB Web site. This will fill a major
gap in the environmental toxicology community, as very limited resources in this area exist in
the public domain. Such information should have high utility in building and interpretation of
predictive toxicology models. In addition, researchers and regulatory scientists can access the
toxicity information to address numerous questions specific to hazard identification and
characterization of environmental chemicals, along with retrospective analyses that will direct
and evaluate possible changes to toxicity studies guidelines.
Synopsis: Thirty years of registration toxicity data and open literature studies have been stored
historically as hardcopy and scanned documents by EPA and others. A significant portion of
these data, including chronic, cancer, developmental, and reproductive studies from laboratory
animals, has now been processed into standardized and structured toxicity data within EPA's
ToxRefDB. These data are now accessible and mineable within ToxRefDB and are serving as a
primary source of validation for EPA's ToxCast™ research program in predictive toxicology. In
addition to providing reference toxicity information to research efforts, ToxRefDB will be mined
for information on the role and impact of previous and current study guidelines on the regulation
of environmental chemicals. The initial collection of studies in March 2006 focused primarily on
the reviews of registrant-submitted toxicity studies on pesticide active ingredients. ToxRefDB
design, development, and implementation were completed in mid-FY2006, with ongoing
updates to the standardized vocabulary and data entry tool interface. The entry of more than
2,000 studies spanning the majority of the ToxCast™ Phase I chemical set was completed late
FY2008. The statuses of these initial datasets are either published, in press, or submitted and
are being made publicly available through the ToxRefDB Web site. Over the next year, a Web-
based query tool for the entire contents of ToxRefDB will become available to the public, and
45
-------
EPA CompTox Research Program FY2009 to 2012
Final
this will be performed in conjunction with a quarterly update of the EPA ACToR program.
ToxRefDB will continue to enter available data from chronic, cancer, developmental, and
reproductive studies, with a focus on potential ToxCast™ Phase II chemicals. The availability
and entry of toxicity data into ToxRefDB also will guide the selection of ToxCast™ Phase II
chemicals. Over the next year, ToxRefDB also will be expanded to capture DNT study data and,
possibly in vivo data submitted through the Agency as part of the EDSP. The completion of
retrospective analyses on reproductive toxicity studies will be completed in FY2010, and the use
of ToxRefDB for additional analyses including rat and mouse chronic/cancer studies, and rat
and rabbit development study assessments will also be undertaken during the period of this
implementation plan.
Partnerships/Collaborations (Internal & External):
(1) EPA ToxCast™ program—provide the reference toxicity information for interpreting the
screening data with respect to animal toxicity information
(2) Tox21 partnership—provide the reference toxicity information for interpreting the screening
data with respect to animal toxicity information
(3) EPA ACToR program—provide ACToR with ToxRefDB data for chemical indexing and
searchability
(4) EPA centers, offices, and laboratories (OPPT/OPP/OSCP/NHEERL)—provide legacy
toxicity data; searchable and computable toxicity data to offices; and entry of additional
study types and from various sources
(5) OECD (including BfR, RIVM, PMRA)—evaluation of current testing guidelines and
assessment of proposed new guidelines (e.g., extended one-generation reproduction
toxicity study)
Milestones/Products:
FY09
(1) Publication on ToxRefDB
(2) Release of stand-alone ToxRefDB data entry tool
(3) ToxRefDB Web page online
(4) Initial public release of selected chronic/cancer end points
(5) Public release of selected reproductive toxicity end points
(6) Public release of selected developmental toxicity end points
(7) Collection of ToxCast™ Phase II chemical toxicity data
(8) Public release of ToxRefDB Web-based query tool
(9) Complete entry of targeted set of chemicals and study types for Phase II of ToxCast ™
(10) Complete reproductive toxicity study retrospective analysis
FY10
(1) Quarterly releases with new data in conjunction with ACToR
(2) Implementation of a process to gather and enter open literature studies
(3) Expansion of ToxRefDB to capture DNT studies and EDSP data
(4) Complete retrospective analyses on other major study types
(5) Release of ToxRefDB live data entry tool
FY11
Quarterly releases with new data in conjunction with ACToR
FY12
Quarterly releases with new data in conjunction with ACToR
Keywords: computational toxicology; ToxRefDB; ToxCast™; ACToR; database
46
-------
EPA CompTox Research Program FY2009 to 2012
Final
QA Project Plan: Category II. ToxRefDB development is guided by a series of standard
operating procedures. These govern all aspects of the project, including data acquisition,
formatting, quality control and assurance, data entry and maintenance, and system
administration. All QA plans are archived in the EPA internal QA system.
d. ChemModel—The Application of Molecular Modeling to Assessing Chemical Toxicity
Lead/Principal Investigator: James Rabinowitz
Research Issue/Relevance: Insufficient experimental information exists for the evaluation of
the potential of a large number of environmental chemicals to cause toxicity and other
environmental effects. Where data does exist, often it is not ideal for this task. The Agency often
must make decisions about specific chemicals when lacking an ideal data set. Molecular
modeling approaches provide an approach for estimating relevant missing information. One
approach to this problem is to estimate the relevant missing information by extrapolation from
existing information on the chemical of interest and other similar chemicals, making use of
molecular modeling approaches. Knowledge of the mechanisms of toxicity provides a rational
basis for application of these computational tools. The results of these models may be used in
conjunction with experimental information to inform decisions about the relevant chemical and to
prioritize the requirements for obtaining missing experimental data.
Purpose/Objective/Impact: The overall objective of this research is to develop an approach
(including the necessary tools) for the application of molecular modeling methods to Agency
problems, particularly problems resulting from the requirement to make preliminary decisions
about chemicals in a data-poor environment. This includes the preliminary evaluation of
chemical toxicity and the prioritization of chemicals testing needs. Knowledge of the potential
mechanisms of toxicity provides a rational basis for extrapolation from existing information to
derive information about chemicals for which little data exist. The differential step in many
mechanisms of toxicity may be generalized as the interaction between a small molecule (a
potential toxicant) and one or more macromolecular targets. (The small molecule may be the
chemical itself or one of its descendants). Using modern molecular modeling methods
developed for the discovery of novel pharmaceutical agents, it is possible to computationally
predict these toxicant-biomolecular target interactions using a combination of direct computer
modeling of atomic interactions between the toxicant-target pair and correction factors derived
from experimentally derived interactions with similar targets. To employ this approach, a library
of computational models of relevant biomolecular targets is being developed. Molecular
modeling approaches then may be used to interrogate this library for the capacity of specific
environmental molecules to interact with each target. These approaches were developed for the
discovery of new pharmaceuticals, where the objective is to discover molecules that interact
most potently with the target. However, the Agency's need is to discover whether chemicals of
environmental interest interact with the target, even if their interaction is much weaker than seen
with potential pharmaceutical agents. An objective of this research is to evaluate the relevant
molecular modeling methods in relationship this Agency requirement.
The endocrine system provides a test for the utility of this approach because many of the
pathways for toxicity and the macromolecular targets in those pathways have been identified.
Appropriate experimental crystal structures of many of the receptor protein targets are available
to create the computational library of targets. Additionally, experimental data of the capacity of a
library of chemicals to displace the natural ligand from the rat estrogen receptor is available
from a single source. Although most of the chemicals in this library have been shown not to
interact with the receptor in this laboratory assay, the few that displace estrogen are orders of
47
-------
EPA CompTox Research Program FY2009 to 2012
Final
magnitude less potent than the natural ligand. The data from this chemical library provides an
opportunity to test the toxicant-target approach and its capacity to separate less potent
chemicals from a large number of similar inactive chemicals. In addition, this exploration
addresses the specific Agency need to evaluate the potential of chemicals to disrupt the
endocrine system. A similar approach may be used to investigate other health effects and fate
and metabolism.
The ultimate objective of this research is to develop a library of biomolecular targets for
chemical toxicity and the methods appropriate for their application to predicting the capacity of a
chemical to interact with these targets. This library of targets may then be used in conjunction
with other approaches as part of a chemical prescreen.
Synopsis: The differential step in many mechanisms of chemical toxicity may be generalized as
the interaction between a small molecule (a potential toxicant) and one or more macromolecular
targets. The small molecule may be the chemical itself or one of its descendents. Describing the
potential of a molecule to participate in interactions of this type is a source of insight chemical
toxicity. In this project, a series of molecular models (148) for critical toxicity targets is being
developed, and methods to evaluate the capacity of a small molecule to interact with these
targets assessed. These methods are adopted from those used in the design of novel
pharmaceutical. A study of a library of 280 environmental chemicals interacting with the
estrogen receptor target is in the final stages of completion. In this library, 14 of the chemicals
are weakly active (3 to 5 orders of magnitude less active than estrogen), and the others are
inactive. Modeling the potential interaction of these chemicals with the rat estrogen receptor
provides an ordered list of molecules. The best results are achieved using a pharmacophore
filter. With that approach, all 14 active chemicals are identified in the first 22 chemicals. In
addition to the importance of these results relative to potential binding to the environmentally
important estrogen receptor, they indicate that this approach may be used to find chemicals that
interact weakly with the target. All 150 of the targets have been interrogated with the ToxCast™
chemicals. Based on the results from the estrogen receptor study, pharmacophores for as many
of the targets as possible will be developed. The analysis of this data will proceed by
comparison with specific ToxCast™ end point data and in concert with short-term data to
evaluate more complex biological end points. The logical extension is to consider the androgen
receptor, where relevant data for comparison and developing a pharmacophore are available.
When sufficient data on other biological macromolecules that are relevant to the Agency
requirements become available, the current library of targets will be expanded. This library of
targets will be used to study chemicals and families of chemicals of importance to the Agency.
The toxicant-target approach described above models molecular identification processes. In
collaboration with other EPA scientists, similar approaches are being applied to the steps that
follow identification. A study on the differential metabolism of pyrethroids and the effect of
stereo-structure on biological clearance is underway, as is a study of perfluorinated chemicals.
Partnerships/Collaborations (Internal & External): Scientists from NHEERL/RTD have
provided the database of the interactions of molecules with the estrogen receptor. We continue
to interact with them relative to this data and the biological details of the computational modeling
effort. Scientists from NERL/HEASD are collaborators on the study of the metabolism and fate
pyrethroids. Collaboration with CDC scientists relative to using the target-toxicant approach to
investigate the interaction of environmental chemicals with nervous system enzymes and
receptors is being developed. As described above, these studies apply methods developed for
pharmaceutical discovery to model the capacity of an environmental chemical to interact with a
macromolecular target.
48
-------
EPA CompTox Research Program FY2009 to 2012
Final
Milestones/Products:
FY09
(1) Report on the capability of the target-toxicant paradigm to identify chemicals that bind
weakly to the estrogen receptor, including a description of the method
(2) Description of the library of 148 biological macromolecule targets
(3) Report on molecular modeling studies of the potential biological effects of the perfluoro
compounds
FY10
(1) Report on the metabolism of pyrethroids and the effects of three dimensional chemical
structures
(2) Description of additional targets added to the target library
(3) Report on the interaction of ToxCast™ chemicals with nuclear receptor targets and the
importance of pharmacophore filters
FY11
(1) Report on the integration of results from the target library and available experimental
parameters
(2) Report on the comparison of results with and without pharmacophore filter for ToxCast™
chemicals
FY12
(1) Comparison of results using the target library with experimentally determined activities,
particularly when observed at the molecular level
(2) Report on the potential use of molecular modeling and the target toxicant paradigm for
regulatory purposes, including a discussion of the OECD principles either as they currently
exist or relative to molecular modeling specific principles
Keywords: molecular modeling, protein binding, toxicity prescreening, weak interactions
QA Project Plan: Category IV. The quality objectives for molecular modeling are to achieve the
best balance between reasonable computational speed and model performance. Development
is guided by a series of standard operating procedures. These govern all aspects of the project,
including data acquisition, formatting, quality assurance, database filling and maintenance, and
system administration. All QA plans are archived in the EPA internal QA system.
e. ToxCast™— Screening and Prioritization of Environmental Chemicals Based on
Bioactivity Profiling and Predictions of Toxicity
Lead/Principal Investigator: Keith Houck
Research Issue/Relevance: The objective of the ToxCast™ research program developed by
NCCT of EPA's ORD is to develop cost-effective innovative approaches to efficiently screen and
prioritize a large number of chemicals for toxicological testing. Using data from state-of-the-art
HTS bioassays developed by the pharmaceutical industry, ToxCast™ is building computational
models to predict the potential human toxicity of chemicals. These hazard predictions should
provide the Agency's regulatory programs with science-based information that will be helpful in
setting priorities for more targeted toxicological evaluations that will help the Agency focus on
those chemicals and end points with the greatest potential for causing adverse effects in
humans. The ultimate goal of ToxCast™ is to deliver an affordable, efficient, science-based
system for categorizing chemicals according to their predicted toxicities.
An essential component of the ToxCast™ research program is the development of a
standardized, reference database containing animal toxicity studies called ToxRefDB.
49
-------
EPA CompTox Research Program FY2009 to 2012
Final
ToxRefDB is being populated with the results of guideline animal toxicity studies on pesticidal
active chemicals that are submitted to the Agency by manufacturers as a requirement of
licensing a pesticide product. ToxRefDB is, for the first time, providing a searchable, mineable
historical database for accessing a wealth of reference in vivo study data. Most importantly,
ToxRefDB will provide the essential interpretive context to anchor ToxCast™ in vitro data (i.e.,
HTS and genomic data) to animal toxicity endpoints with selected ToxRef in vivo outcomes
serving as the basis for developing predictive in vitro bioactivity profiles and signatures. Equally
essential to the overall success of this project will be the development of a suitable informatics
and analysis infrastructure for storing, relating, and extracting patterns from all data associated
with the ToxCast™ project, including chemical, HTS, and in vivo data elements.
ToxCast™ databases and predictive models for the potential toxicity of environmental chemicals
will be useful to EPA program offices for chemical prioritization. For example, OPP and the
OPPT anticipates taking advantage of ToxCast™ models and datasets to prioritize in vivo
animal testing of products that have limited toxicity data available such as:
• antimicrobial pesticides,
• inert ingredients in pesticide products,
• manufacturing process impurities,
• metabolites and environmental degradates of concern, and
• new and existing industrial chemicals.
After internal clearance and external peer review, the information in ToxRefDB and HTS data
generated on the chemicals screened in ToxCast™ will be publicly available at
www.epa.gov/ncct/toxrefdb and www.epa.gov/ncct/toxcast.
Purpose/Objective/Impact: The objective of the ToxCast™ research program is to develop a
cost-effective and rapid approach for screening and prioritizing a large number of chemicals for
toxicological testing (Dix et al., 2007). Using data from HTS bioassays developed in the drug
discovery community, ToxCast™ is generating data, constructing databases, and building
computational models to forecast the potential human toxicity of chemicals. These hazard
predictions should provide EPA regulatory programs, including OPP, with science-based
information helpful in prioritizing chemicals for more detailed toxicological evaluations, ultimately
leading to reduced animal testing. ToxCast™ is currently in the PoC phase, wherein more than
300 chemicals have been assayed in more than 600 different HTS bioassays, creating rich
bioactivity profiles for these chemicals. The Phase I chemicals are primarily conventional
pesticide actives that have been extensively evaluated using traditional mammalian toxicity
testing and, hence, have a number of well characterized toxicity outcomes (e.g., carcinogenicity;
and developmental, reproductive, and neural toxicity). These in vivo data, in turn, have been
extracted from the evaluations conducted by OPP scientists and were used to construct and
populate ToxRefDB. Comparable toxicity data from other toxicity sources (e.g., NTP) also are
being captured in ToxRefDB. A broader and more diverse set of complementary data on
thousands of chemicals is being identified and collated in EPA's ACToR. ACToR (as well as the
analysis component, ToxMiner) is providing the essential informatics infrastructure for housing,
integrating, and analyzing all chemical and assay data associated with the ToxCast™ project,
also in the context of a much larger world of Web-accessible chemical-toxicological information.
DSSTox, in turn, is providing high-quality, standardized chemical structure indexing for ACToR
and ToxCast™, including other high-interest Agency chemical data inventories.
ToxRefDB is critical to developing predictive signatures, because it links ToxCast™ HTS in vitro
data to in vivo toxicity end points associated with the same chemicals. The toxicity data in
ToxRefDB and the HTS data generated in ToxCast™ will be made publicly available through
50
-------
EPA CompTox Research Program FY2009 to 2012
Final
EPA Web sites and databases. The first manuscript on ToxRefDB recently was published
(Martin et al., 2008), presenting toxicity profiles from 2-year rodent bioassays on 310 chemicals.
A similar analysis is nearing completion on multigeneration reproduction and prenatal
developmental test data for the ToxCast™ chemicals in ToxRefDB, profiling the toxicity potential
of this chemical set across generation, life stage, and different classes of end points.
ToxCast™ is a multiyear effort that is divided into three distinct phases:
• Phase I: 300 chemicals assayed in more than 600 different HTS bioassays, to create
predictive bioactivity signatures based on the known toxicity of the chemicals;
• Phase II: focused on confirmation and expansion of ToxCast™ predictive signatures,
generating HTS data for at least 300 additional chemicals; and
• Phase III: ToxCast™ expanded to thousands of environmental chemicals for which little
toxicological information is available.
Once ToxCast™ has gone through successful initiation of Phase III, the data and toxicity
predictions will be ready for deployment throughout numerous EPA program offices. NCCT will
work to link these hazard predictions with exposure predictions and create integrated database
analysis tools facilitating customized chemical prioritizations appropriate to specific programs.
Synopsis: The primary goals for ToxCast™ are the completion of Phase I data collection and,
concurrently, development of the informatics infrastructure to store and analyze these data;
derivation of predictive signatures from Phase I data and validation of these signatures with
Phase II data; and the application of these predictions to the prioritization of chemicals in
various chemical and nanomaterial testing programs. Success in meeting these four principal
goals will lead to secondary applications in developing human toxicity pathway analyses, and in
high-throughput risk assessments.
Partnerships/Collaborations (Internal & External):
• Tox21 with NTP and NCGC
• OECD Molecular Screening Project
• OECD Enhanced One-Generation Reproductive Test Guideline
• EPA/ORD/NHEERL DNT Team
• EPA/ORD/NCEA Phthalate Team
• MOUs and MTAs with over 20 external organizations collaborating on ToxCast™ assays,
chemicals, and data analysis.
Milestones/Products:
FY09
(1) Completion of ToxCast™ Phase I data collection
(2) Provide annotated Phase I datasets for public access
(3) Derivation of predictive signatures from ToxCast™ Phase I data
(4) Multiple publications on ToxCast™ datasets
(5) Publication on signature generation (SCOUT item, April 2009)
(6) Publication on analysis of NR pathways and toxicity
(7) Convene first ToxCast™ Data Summit for identifying promising prediction models from
intramural and extramural sources
(8) Finalize selection of chemicals for next 3 to 5 years of ToxCast™ and Tox21 projects
(9) Prioritize and order 4000 to 6000 chemicals in collaboration with other Tox21 partners
FY10
51
-------
EPA CompTox Research Program FY2009 to 2012
Final
(1) Publications describing approaches to combining exposure, PK, and in vitro assays to do
risk prioritization
(2) Evaluate compatibility of nanomaterials of diverse classes with ToxCast™ assays
(3) Prioritize and select assays to be run for ToxCast™ Phase II
FY11
(1) Confirmation of ToxCast™ predictive signatures with Phase II data
(2) Publications on signature confirmations and applications
(3) Completion of ToxCast™ Phase II data collection
FY12
Application of ToxCast™ predictions to the prioritization of chemicals in various EPA chemical
and nanomaterial testing programs
Keywords: ToxCast™, high-throughput screening, hazard, predictive toxicology, chemical
prioritization
QA Project Plan: Category I. ToxCast™ Quality Management Plan, includes: Information
Management QA Project Plan, NCGC QA Project Plan (for IAG), and nine separate contractor
QA plans and records. All QA plans are archived in the EPA internal QA system.
f. ExpoCast™—Exposure Science for Screening, Prioritization, and Toxicity Testing
Lead/Principal Investigator: Elaine Cohen Hubal
Research Issue/Relevance: High-visibility efforts in toxicity testing and computational
toxicology raise important research questions and opportunities for exposure scientists. There is
a clear need for a collaborative effort across the exposure and risk assessment community to
ensure that the required exposure science, data, and tools are ready to address immediate
needs resulting from application of high-throughput in vitro technologies for toxicity testing. A
coherent program is required to formulate significant exposure questions posed by these novel
in vitro toxicity data; develop creative approaches for applying existing exposure information and
tools to address these questions; and, finally, identify key exposure research needs to interpret
the toxicity data for risk assessment. The authors of the National Academies report (NRC, 2007)
emphasize that population-based data and human exposure information are required at each
step of their vision for toxicity testing, and that these exposure data will continue to play a critical
role in both guiding the development and use of the toxicity information. Exposure research
questions posed in this report include how to (1) use information on host susceptibility and real-
world exposures to interpret and extrapolate in vitro test results, (2) use human exposure data
to select doses for toxicity testing so information on biological effects pertains to environmentally
relevant exposures and (3) relate human exposure data from biomonitoring surveys to
concentrations that perturb toxicity pathways to identify biologically relevant exposures. NCCT
has identified the need to include exposure information for chemical prioritization, modeling
system response to chemical exposures across multiple levels of biological organization
(through to the population level), and linking information on potential toxicity of environmental
contaminants to real-world health outcomes (Cohen Hubal et al., 2008). Together, scientists
from NCCT and NERL's Human Exposure Research Program have identified and will be
conducting research in the following priority exposure research areas to support chemical
screening, prioritization, and toxicity testing (Sheldon and Cohen Hubal, 2009): (1) accessible
and linkable exposure databases, (2) exposure screening tools for accelerated chemical
prioritization, (3) computational tools for dose reconstruction and source-to-outcome analyses,
(4) tools for understanding the fundamental processes and factors influencing human
exposures, and (5) efficient monitoring methods to measure and interpret biologically relevant
52
-------
EPA CompTox Research Program FY2009 to 2012
Final
exposure metrics. Susceptibility, vulnerability, and life-stage aspects are integral to each of
these research areas.
Purpose/Objective/Impact: The ExpoCast™ program is being initiated in FY09 to ensure the
required exposure science and computational tools are ready to address global needs for rapid
characterization of exposure potential arising from the manufacture and use of tens of
thousands of chemicals and to meet challenges posed by new toxicity testing approaches. The
overall goal of this project is to develop novel approaches and tools for screening, evaluating,
and classifying chemicals based on the potential for biologically relevant human exposure, to
inform prioritization and toxicity testing. An emphasis will be placed on conducting research to
mine and translate scientific advances and tools in a broad range of fields to provide information
that can be used to support enhanced exposure assessments for decisionmaking and improved
environmental health. Advanced exposure databases, computational tools, and analysis
approaches are required to prioritize chemicals, to design effective in vitro screening protocols,
and to interpret the results of these screening tests for human health risk assessment.
Approaches for integrating information on genetic susceptibility and life-stage and population-
level vulnerabilities with in vitro toxicity data also are required to improve public-health
decisionmaking. This initiative will advance Agency tools for efficiently characterizing and
classifying chemicals based on potential for biologically relevant exposures. The improved
exposure science and knowledge will subsequently inform the characterization of
environmentally-relevant toxicity.
Synopsis of NCCT Directed Research: ExpoCast™ will provide an overarching framework for
the science required to characterize biologically relevant exposure in support of the Agency
computational toxicology program. Broadly and long term, the ExpoCast™ program will foster
novel exposure science research to (1) inform chemical prioritization, (2) understand the
systems response to chemical perturbations resulting from environmentally relevant exposures
and how these translate to relevant biological changes at the individual and population levels,
and (3) link information on potential toxicity of environmental contaminants to real-world health
outcomes. The NCCT-directed aspects of the ExpoCast™ program will have a strong focus on
research required to interpret and translate in vitro hazard data in the context of real-world
exposures for risk assessment. Research will be conducted jointly with NERL to leverage
expertise and resources required to meet objectives of this multidisciplinary project.
An important early component of ExpoCast™ will be to consider how best to consolidate and
link human exposure and exposure factor data for chemical prioritization and toxicity testing.
Under the ExpoCast™ program, NCCT and NERL scientists will collaboratively
• evaluate and recommend approaches for improving accessibility to EPA human exposure and
exposure factor data and for facilitating links between exposure and toxicity data (e.g.,
through DSSTox and ACToR systems),
• advocate for the creation of a consolidated EPA exposure database focused on measured
and predicted concentrations in exposure and biological media, and
• propose standards for human exposure data representation.
Early research activities will focus on identifying and evaluating novel approaches for
characterizing exposure to prioritize chemicals and developing modeling approaches for
considering exposure potential based on chemical properties, sources (e.g., consumer
products), uses, life cycle, and individual and population vulnerability. Specific tasks pertinent to
these goals include:
• analysis of extant exposure data to identify the critical metrics and develop simple indices for
representing biologically-relevant personal exposure over time, place, lifestage, and lifestyle
53
-------
EPA CompTox Research Program FY2009 to 2012
Final
or behavior.
• development of novel approaches for characterizing biologically-relevant exposure to
prioritize chemicals, some examples follow:
- application of residential models for prioritizing SVOCs,
- development of dermal uptake model suite,
- application of biomonitoring equivalent (BE) approach to interpret ToxCast™ data;
• development of human exposure knowledgebase; and
• application of genomic tools and other biomarkers of exposure and susceptibility to consider
population-level vulnerabilities for toxicity testing and risk assessment.
NCCT Partnerships/Collaborations (Internal & External): The ExpoCast™ project will be
conducted in close collaboration with the ToxCast™ program and through extensive
collaboration with NERL principal investigators in the Human Exposure Research Program.
Integration of research directed by the both NCCT and NERL will be enhanced by activities
such as the Computational Toxicology Rotational Fellow Program. To jump start joint research
activities, a NERL investigator will join the rotational fellows program in the summer of 2009 to
focus on improving access to human exposure data. Partnerships with the other laboratories
and centers also will be developed as the ExpoCast™ program advances. For example,
collaboration with NHEERL on the MICA study is providing the opportunity to pilot advanced
computational approaches for evaluating multifactorial biomarker data (including genomic data)
across the exposure-outcome continuum to investigate the interplay of environmental and
genetic factors on complex diseases.
External collaborations include research with Dr. John Little of Virginia Tech, to develop
improved tools for rapidly predicting exposure associated with SVOCs used or emitted in the
residential environment. Dr. Sean Hays and coworkers of Summit Toxicology will be exploring
application of the BE approach for interpretation of ToxCast™ data. In a collaboration
established through Bio-chem Redirect Program and implemented through the International
Science and Technology Center (ISTC), Dr. Petr Nikitin of the Natural Science Center (NSC) of
A.M. Prokhorov General Physics Institute, Russian Academy of Sciences is leading research to
develop a multichannel immunosensor for detection of pyrethroids. The primary goal of this
project is development of a biochip technology that avoids sophisticated labeling steps. This is
an example of the type of research required to address needs for advanced exposure
monitoring tools. In collaboration with the ICCA-LRI, we are participating in planning of a
workshop focused on developing innovative tools to characterize biologically relevant
environmental exposures and implication of these for health risks. Finally, in collaboration with
the Environmental Bioinformatics STAR center at UMDNJ, we will be presenting a symposium
at the ISES 2009 annual meeting.
Input to and feedback on ExpoCast™ will be solicited through the ExpoCoP to facilitate
integration and collaboration across the broader scientific community. ExpoCoP includes
representatives from the ORD laboratories and centers, Agency program offices, other Federal
government agencies, academia, industry, and environmental advocacy groups. International
representatives also participate in the ExpoCoP.
Milestones/Products:
FY09
(1) EHP paper with NERL, "Exposure as part of a systems approach for assessing risk"
(2) Tox. Sci. Forum paper "Biologically Relevant Exposure for Toxicity Testing"
(3) ExpoCoP monthly teleconference, ESC resource, face-to-face meeting at ISES 2009
(4) ICCA-LRI workshop, "Connecting Innovations in Biological, Exposure and Risk Sciences:
Better Information for Better Decisions"
54
-------
EPA CompTox Research Program FY2009 to 2012
Final
(5) SVOC workshop, "Semi-Volatile Organic Compounds (SVOCs) in the Residential
Environment"
(6) ISES 2009 Annual Conference, "Transforming Exposure Science for the 21st Century"
(7) Symposium at ISES 2009, "Integrative Exposure Biology and Computational Toxicology for
Risk Assessment"
(8) Survey and identify high-priority exposure data resources for initial chemical indexing in
collaboration with ACToR and DSSTox
(9) ExpoCast™ conceptual framework and research plan
FY10
(1) Workshop to review and evaluate current exposure prioritization tools
(2) Position paper recommending standards for exposure data representation
(3) White paper defining exposure space and plan for assessing exposure data landscape
(4) White paper exploring development and application of human exposure knowledgebase
(5) Begin implementation of standards across exposure databases of highest interest and utility
for NCCT projects
FY11
(1) Manuscript describing extant data analyses to identify critical determinants for exposure
classification and chemical prioritization based on potential for exposure
(2) Guide incorporation and further development of simple exposure estimation tools within the
ACToR system for use in prioritization
FY12
Apply exposure index for prioritization to subset of ToxCast compounds to evaluate concept.
Keywords (three to five): exposure science, chemical prioritization and toxicity testing,
vulnerable populations, susceptibility
QA Project Plan: Category III. QA of modeling projects serves at least two overlapping goals:
(1) Verification - Reproducibility of results is essential for the scientific method, and (2)
Continuity Proper documentation of results allows future researchers (or the same researcher
after a long period of time) to return to a project without excessive amounts of time spent
understanding what was done before. All QA plans are archived in the EPA internal QA system.
g. v-Embryo™—Virtual Embryo
Lead/Principal Investigator: Thomas B. Knudsen, PhD
Research Issue/Relevance: Research Issue: The Virtual Embryo project (www.epa.gov/ncct/v-
Embryo/) is motivated by scientific and regulatory needs to understand mechanisms of
developmental toxicity. A key research issue is to model how the embryo reacts to
environmental chemicals as a "complex system." Navigating this complexity requires detailed
knowledge of molecular embryology, data on cellular systems using HTS approaches, and
computational models of network dynamics and multicellular function.
Morphogenesis entails a dynamic tissue flow that is driven by conserved cell signaling pathways
and cellular reaction networks that follow these pathways during stimulus, mutation, or injury [1],
Our strategy is to modularize the embryo as a collection of tractable models that represent the
cell as a computational unit [2], In this strategy, the cell is treated as an autonomous agent that
processes local signals and selects from a repertoire of core behaviors that include growth,
differentiation, mitosis, apoptosis, migration, adhesion, and cell-shape changes. Specific rules
for signal-response are programmed for molecular pathways, cellular dynamics, and unique
55
-------
EPA CompTox Research Program FY2009 to 2012
Final
biology per morphogenetic system [3], Sophisticated imaging techniques can reveal complex
dynamics of cell-cell relationships and a "morphogenetic blueprint" of early development [4],
Although much is known about molecular signaling networks that drive morphogenesis,
considerably less is known about the nature of "higher-order" processes that control collective
cellular behavior [3], In complex systems, molecular networks can invoke higher level processes
through signal-input and response-output relationships, as determined by the timing and
function of signal strength and dynamic range [5], We take this hypothesis in the context of
environmental stressors to embryonic development. Understanding network-state relationships
will be required to predict nonlinear dose-response relationships and when a breakdown of
higher order control systems may occur [6], Cell-based computational models have been used
to predict emergent properties that arise from cooperative transactions of cell groups behaving
as a self-regulating system [7], Homeostasis, adaptation, and repair are a few examples of
emergence in a perturbed system [8],
Relevance: A new strategy for developmental toxicity testing involves screening multiple
chemicals through cell-based in vitro assays [1], The goal is to build robust signatures of toxicity
that translate into in vivo predictions [9], Because tissues are more than a cumulative sum of
individual cell behaviors, computational models can accelerate this effort [7], A "virtual tissue"
(VT) representation reconstructs a broad range of biological responses following cell-based
rules and systems-level controls. VT models rapidly can sweep parameter space following
chemical or genetic perturbation to predict aggregate cellular behaviors and higher order
responses [10],
Purpose/Objective/Impact: Purpose: EPA's Virtual Embryo (v-Embryo™) will comprise a
knowledgebase (VT-KB) of relevant information and simulation engine (VT-SE) on the front end
of the modeling software. Proof-of-principle is underway to explore the range of potential
applications where comprehensive simulation is reliable and to find the limit in scale or stages
where studying the real embryo is not cost effective. Computational models have been built by
others for in silico reconstruction of chondrogenesis [11], gastrulation [12], angiogenesis [13],
and somitogenesis [14], These models were implemented as hybrid cellular automata using
CC3D open-source tissue-simulation environment [www.CompuCell3D.org], which is being
evaluated for the Virtual Embryo. Initial models will focus on specific morphogenetic systems
that replay important concepts in experimental embryology and that are targets in
developmental toxicity. The end goal is a library of computer-driven simulations that can be
manipulated in silico and correlated with in vitro responses or in vivo phenotypes to predict
developmental toxicity. Applications for developmental toxicity align with EPA's strategy on the
future of toxicity testing [15] and can leverage unique pathway-based data for numerous
chemicals tested in mouse embryonic stem cells (mESC), free-living zebrafish (ZF) embryos,
and ToxCast™ assays [16] to
simulate key signaling pathways, interlocking genetic networks and cellular dynamics in
developing tissues;
model how embryonic cells react as agents to chemical exposure individually and collectively as
a complex system;
analyze emergent behaviors and canalizing influences following stimulus, injury, and
perturbation; and
understand how this complexity contributes to the differential susceptibility of embryonic tissues
across chemical, dose, stage, genetic makeup, and time.
Objective: The main objective is to build data-rich models that can be used to analyze causal
relationships during environmental and genetic perturbation. One initial prototype will focus on
56
-------
EPA CompTox Research Program FY2009 to 2012
Final
the powerful sine oculis network that controls early eye development - conserved
morphogenetic pathways [17], patterns of malformation and sensitivity to chemical perturbation
[18], There are many advantages of focusing on eye malformations and, specifically, the
molecular events related to Pax6 leading to this end point: strong knowledgebase, relevant
human phenotypes, and underlying genetic susceptibility and environmental sensitivity. The
underlying molecular pathways and signaling networks scale to tissue-level developmental
effects in many different systems. A second prototype will focus on the patterning systems
controlling early limb-bud development. The specific objectives are as follows.
1. Build knowledgebase (VT-KB) and front-end simulation engine (VT-SE) for developmental
processes and toxicities.
Rationale: VT-KB is required to initially store gene-gene and gene-phenotype associations
that will be used to provide the rules for cell-based modeling in the VT-SE. The framework
will rely on information from the literature, data from national repositories (GO, EMAGE,
GXD, MPO, ZFIN, and OMIM) and pathway analysis software. VT-KB design, development,
and implementation are supported through ITS-ESE contract no. 68-W-04-005, task order
no. 058: Technical Support for Development of Developmental Systems Toxicity Network
(DevToxNet). Perl-scripts written for information extraction and assisted curation of relevant
facts from scientific literature returned an output matrix loaded to MySQL. Initial application
will build gene-gene and gene-phenotype associations during eye and limb development for
rat, mouse, zebrafish, or human species. VT-SE comprises a front end to the modeling
software (DDLab, CC3D, C++, Python, Blender3D, and GanttPV). An interactive tool is
being developed with support from the Environmental Modeling and Visualization Laboratory
under ITS-ESE contract no. 68-W-04-005, task order no. 02: Virtual Embryo: Simulation and
Visualization Project Management Plan. Initial application of VT-SE is to construct a cell-
based, network-driven model for lens-retina induction that reconstructs the cellular dynamics
and morphogenetic blueprint of ocular dysmorphogenesis. This model can be quantitative in
terms of the degree of severity of chemical-induced defects (graded dose response) and
relative risk for incidence of responding embryos (quantal dose response).
2. Construct cell-based computational models for prototype morphogenetic processes and
embryonic modules.
Rationale: The second specific aim uses the VT-SE to model modular embryonic systems
and specific morphogenetic events and their perturbation. The strategy will apply agent-
based models (ABMs). Initial prototypes (optic cup and limb bud) have well-characterized
signaling networks and differential susceptibility to chemical disruption; other systems will be
added over time. Both small prototypes are organized by complex self-regulating networks
of signal molecules commuted from cell-signaling centers and described mathematically as
Turing gradients. Prevailing models entail reciprocal induction in which heterotypic
interactions between presumptive lens epithelium and prospective neural retina lead to
formation of the optic cup, and interactions between apical ectodermal ridge and underlying
mesenchyme drive polarized outgrowth of the paddle-shaped limb bud. Both processes are
organized by a self-regulating network of genes and signaling gradients (FGFs, BMPs,
SHHs). Although the dual-reciprocating models set the stage for emergence of the optical
neuraxis and appendicular skeleton, respectively, they do not explain higher order
processes that control geometry and size of these rudiments, nor do they account for
differential susceptibility to teratogens [18], For this purpose, we propose the extended
cellular large-Q Pott's model (CPM) implemented in CC3D [19] and managed with Python
software as part of the VT-SE.
57
-------
EPA CompTox Research Program FY2009 to 2012
Final
3. Specify rules for component interactions of developmental pathways at the cellar and
molecular scales.
Rationale: Network structures for regulatory pathways and cellular systems will be portrayed
using a Boolean (on-off) formalism. A Boolean Network (BN) qualitatively captures system
behavior probabilistically (PBN) or deterministically (DBN); the former is more biologically
plausible, whereas the latter is a modeling tool of the whole process, which enables us to
simulate, analyze, and manipulate different parts of the system. Both models incorporate
rule-based dependencies for gene-gene and cell-cell interactions that can be built with
information from the VT-KB. States of the network, as determined by wiring and rules, will be
characterized using DDLab software, which identifies stable attractors in complex systems
[20], The attractor concept implies that a finite number of stable cell states exist in a complex
system as pathways of differentiation or canalization. Order and timing have great
importance for the predictive modeling of genetic errors or cellular disruptions in the embryo
caused by phase-specific chemical effects. As such, the ability to introduce predefined or
stochastic lesions will enhance the functionality of VT-SE. Virtual Embryo is building a
prototype interface that manages the order and timing of gene expression and signaling
events for Python-based components in the VT-SE. This tool is being developed under ITS-
ESE contract no. 68-W-04-005, task order no. 02: Virtual Embryo—Simulation and
Visualization.
4. Analyze abnormal developmental trajectories predicted from cellular MOAs that follow
chemical perturbations.
Rationale: A high-fidelity computer program that links cellular processes with network-level
function can be evaluated for its capacity to evolve features not explicitly coded in the cell-
based model. As noted above, this emergence is important for manifesting the response to
genetic errors and cellular disruptions that are introduced to the model as targets of
developmental toxicity, based on simulated and experimental data. Such a computational
model can reveal an interaction of mechanisms at the cellular and molecular scales to
produce emergent phenomena that manifest as abnormal developmental phenotypes [10],
Rules will derive from simulated data and semiarbitrary parameters in ocular or related
systems because it will take a major experimental effort to model parameters kinetically for
all relevant pathways, reactants, and interactions. Eventually, such information would be
helpful to build a quantitative model. CC3D software advancements will be needed to
implement molecular motors for core cell behaviors and to parallelize this implementation.
Impact: This research aims to improve mechanistic understanding and predictive modeling of
developmental toxicity. Biological models that are simple enough so as to be computationally
feasible (tractable) and yet complex enough to compute integrated cellular behaviors (rational)
can reveal key events in multicellular organization, classify abnormal developmental trajectories
from genetic network inference, and predict chemical dysmorphogenesis from pathway-level
data. The initial focus on existing data, with use of the modeling effort to identify data gaps and
to help guide the design of experiments for generating additional data as needed, can produce
results that provide significant new information on likely dose-response and time-course
behaviors of developmental toxicants. Most developmental modeling has been qualitative,
showing the link between fundamental processes and morphogenesis. Virtual Embryo is moving
to a different, quantitative level through knowledge of molecular embryology and pathway-level
data from HTS efforts. That resource can have an impact on HTP hypothesis testing (parameter
sweeps) to inform experimental design or to dry-run intractable experiments complicated by
58
-------
EPA CompTox Research Program FY2009 to 2012
Final
time, scale, and cost (monetary and animal). Over the short term, we anticipate the work will
draw greater attention to integrative thinking, the application of computational models to
understand mechanisms, and approaches for uncertainty analysis and understanding how large
uncertainties about parameter values will affect quantitative prediction. Expanding the prototype
models to broader representation of stages, tissues and species will be an intermediate step
toward the more visionary reconstruction of a "Virtual Embryo."
Synopsis: Motivation of the Virtual Embryo is scientific needs to understand mechanisms of
toxicity and predict developmental defects from complex datasets. The research goal is to
simulate embryonic tissues reacting to perturbation across chemical class, system, stage,
genetic makeup, dose, and time. Data input is detailed knowledge of molecular embryology,
high-throughput data from in vitro models, signaling pathways, and cellular phenotypes. Output
models aim for the modular reconstruction of a developing embryo from cell-based models of
morphogenesis and differentiation.
M i lestones/Products:
FY09-10
(1) Project Plan: Category III QAPP
(2) Recruit: postdoctoral fellow
(3) Manuscript: application of VT-KB to analyze ToxRefDB developmental toxicity studies
(4) Model: VT-KB based qualitative (structural) model of self-regulating ocular gene network
(5) Model: VT-SE based cell-based computational model of lens-retina induction
(6) Manuscript: ocular morphogenesis, gene network inference, analysis, and modeling
FY10-11
(1) Project plan: extend lens-retina model to other stages and species
(2) Model: incorporate pathway data from ToxCast™, mESC, and ZF embryos
(3) Manuscript: sensitivity analysis for developmental trajectories and phenotypes
(4) Project plan: integrate with other morphogenetic models (ES cells, Zfish)
FY11-12
(1) Manuscript: test model against predictions for pathway-based dose-response relationship
(2) Manuscript: uncertainty analysis of models for complex systems
(3) Model: computer program of early eye development, using rules-based architecture, cell-
based simulators, and systems-wiring diagrams
Keywords: embryo development, systems biology, computational modeling
QA Project Plan: Category III. The proposed designation for Virtual Embryo is Quality
Assurance Category III. This designation recognizes its origin as a basic research project
(Category IV) that is moving into proof of concept phase (Category III). A Virtual Embryo Quality
Management Plan (QMP) will be constructed as the project moves into the proof of concept
phase.
Phase-I (development): first-generation ABMs based on the small prototype systems of lens
induction and polarized limb outgrowth (2009-10).
Phase-I I (evaluation): sensitivity analysis using data for ToxCast™ chemicals that disrupt eye
and/or limb development or Tox21 assays of relevant signaling pathways (2010-11).
Phase-Ill (expansion): uncertainty analysis of quantitative models that simulate reaction to
perturbation across chemical, system, stage, genetic makeup, dose, and time (2011-12).
59
-------
EPA CompTox Research Program FY2009 to 2012
Final
The anticipated Date of Elevation to Category-ll is 2012. This is based on the premise that
research enabled by these models will reduce uncertainty in risk assessment for prenatal
developmental toxicity through understanding of complex mechanisms of environmental
chemicals and their impact on complex developing systems. We also anticipate a successful
Virtual Embryo in the long-term can reduce the reliance on animal testing for prenatal
developmental toxicity. Many of the 10,000 chemicals with which EPA is concerned do not have
such information available.
References:
1. National Research Council, Committee on Toxicity and Assessment of Environmental Agents
(2007) Toxicity Testing in the Twenty-first Century: A Vision and a Strategy. Washington, DC:
National Academies Press (http://www.nap.edu/catalog/11970.html).
2. Thorne BC, Bailey AM, DeSimone DW and Peirce SM (2008) Agent-based modeling of
multicell morphogenetic processes during development. Birth Defects Res (part C) 81: 344-
353.
3. Lewis J (2008) From signals to patterns: space, time, and mathematics in developmental
biology. Science 322: 399-403.
4. Keller PJ, Schmidt AD, Wittbrodt J and Stelzer EHK (2008) Reconstruction of zebrafish early
embryonic development by scanned light sheet microscopy. Science 322: 1065-1069.
5. Janes KA, Reinhardy HC and Yaffe MB (2008) Cytokine-induced signaling networks prioritize
dynamic range over signal strength. Cell 135: 343-354.
6. Andersen ME, Yang RSH, French CT, Chubb LS and Dennison JE (2002) Molecular circuits,
biological switches, and nonlinear dose-response relationships. Env Hlth Persp 110: 971-978.
7. Knudsen TB and Kavlock RJ (2008) Comparative bioinformatics and computational
toxicology. In: Developmental Toxicology Volume 3, Target Organ Toxicology Series. (B
Abbott and D Hansen, editors) New York: Taylor and Francis, Chapter 12, pp 311-360.
8. Basanta D, Miodownik M and Baum B (2008) The evolution of robust development and
homeostasis in artificial organisms. PLOS Computat Biol 4(3): e1000030.
9. Martin MT, Houck KA, McLaurin K, Richard A and Dix DJ (2007) Linking regulatory
toxicological information on environmental chemicals with high-throughput screening (HTS)
and genomic data. The Toxicologist CD - J. Soc Toxicol 96: 219-220.
10. Andersen T, Newman R and Otter T (2009) Shape homeostasis in virtual embryos. Artificial
Life 15(2): 161 -183.
11. Chaturvedi R, Huang C, Kazmierczak B, Schneider T, Izaguirre JA, Glimm T, Hentschel HE,
Glazier JA, Newman SA and Alber MS (2005) On Multiscale approaches to three-dimensional
modeling of morphogenesis. JR Soc Interface 2: 237-253.
12. Cui C, Yang X, Chuai M, Glazier JA and Weijer CJ (2005) Analysis of tissue flow patterns
during primitive streak formation in the chick embryo. Developmental Biol 284: 37-47.
13. Mahoney AW, Smith BG, Flann NS and Podgorski GJ (2008). Discovering novel cancer
therapies: a computational modeling and search approach. In: IEEE Symposium on
Computational Intelligence in Bioinformatics and Bioengineering. Sun Valley, ID: CIBCB.
14. Glazier JA, Zhang Y, Swat M, Zaitlen B and Schnell S (2008) Coordinated action of N-CAM,
N-cadherin, EphA4, and ephrin B2 translates genetic prepatterns into structure during
somitogenesis in chick. Curr Topics Devel Biol 81:205-247.
15. Future of Toxicity Testing Workgroup, US EPA (2009) The U.S. Environmental Protection
Agency's Strategic Plan for Evaluating the Toxicity of Chemicals
http://www.epa.gov/osa/spc/toxicitytesting/.
16. Chapin R, Augustine-Rauch K, Beyer B, Daston G, Finnell R, Flynn T, Sunter S, Mirkes P,
O'Shea KS, Piersma A, Sandler D, Vanparys P and van Maele-Fabry G (2008) State of the art
in developmental toxicity screening methods and a way forward: a meeting report addressing
60
-------
EPA CompTox Research Program FY2009 to 2012
Final
embryonic stem cells, whole embryo culture, and zebrafish. Birth Defects Res (Part B) 83:
446-456.
17. Chow RL and Lang RA (2001) Early eye development in vertebrates. Annu Rev Cell Dev Biol
17: 255-296.
18. Green ML, Singh AV, Zhang Y, Nemeth KA, Sulik KK and Knudsen TB (2007)
Reprogramming of genetic networks during initiation of the fetal alcohol syndrome. Devel
Dynam 236: 613-631.
19. Cickovski TM, Huang C, Chaturvedi R, Glimm T, Hentschel HGE, Alber MS, Glazier JA,
Newman SA and Izaguirre JA (2005) A framework for three-dimensional simulation of
morphogenesis. IEEE/ACM Trans Comput Biol Bioinfor2: 1-15.
20. Wuensche A (1996) Discrete dynamics Lab (DDLab) Available from [http://www.ddlab.com],
h. v-Liver™ —The Virtual Liver Project
Lead/Principal Investigator: Imran Shah
Research Issue: The Virtual Liver project (http://www.epa.gov/ncct/virtual_liver) is aimed at
providing decision support tools for evaluating chemical-induced adverse liver outcomes across
chemicals, doses, and species using in vitro data. Considering nuclear-receptor (NR)-mediated
liver cancer as an archetypal chronic adverse outcome, we focus on the research issues: Which
molecular circuits and cellular states altered by chemicals lead to cell damage, death, and
proliferation? How are these cellular perturbations propagated across tissues as lesions? How
can we organize this complexity computationally to develop Virtual Tissues?
Two important perturbed cellular phenotypes, or states, in carcinogensis are (1) initiation, in
which chemical mutagens cause DNA damage rendering a cell resistant to apoptosis, inhibition
of cell proliferation; and (2) promotion, in which mitogenic signals persistently stimulate the
initiated cell creating focal proliferation. Increasing evidence suggests that the NR superfamily
mediates rodent hepatocarcinogenesis for a number of environmental chemicals (Butler, 1996).
For example, di(2-ethylhexyl)-phthalate (DEHP) and perfluorooctanoic acid (PFOA) are PPAR-a
activators (Maloney and Waxman, 1999); whereas pesticides like conazoles and
pyrethroidsactivate are either PXR or both PXR and CAR (Kretschmer and Baldwin, 2005). The
role of NR-mediated activity in molecular circuits is being explored actively through genomic
profiling, and the dose-dependence of specific molecular switches is being assayed across
hundreds of environmental chemicals in ToxCast™.
Propagating cellular alterations spatially requires information flow between cells, which normally
occurs on a backdrop of microanatomic spatial zones with heterogenous levels of nutrients and
distinct spatial distribution of intracellular states (Pette and Wimmer 1979; Oinonen et al. 1998).
The microanatomic distribution of xenobiotics causes zonal alterations in cell states (Kato et al.,
2001) that can progress to cell injury and even death. Hepatocyte (HC) death stimulates
neighbouring cells to replicate (regenerative proliferation). Necrotic death also can lead to
Kupffer cell (KC) activation, migration, and release of inflammatory cytokines, which locally can
accelerate cell injury. There is evidence for such HC-KC interactions in PPAR-a mediated
hepatotoxicity and cancer (Rusyn, 1998; Roberts, 2007). Mitogens also have been shown to
disrupt gap junction communication between cells (Krutovskikh et al., 1995), which can reduce
their homeostatic capacity. Advanced imaging, histomorphometry (HMP), and molecular assays
are making it feasible to extract local information on cells in a microanatomic context.
To computationally model this level of biological complexity requires some simplifying
assumptions about the modular organization of physiologic events across scales. We
61
-------
EPA CompTox Research Program FY2009 to 2012
Final
hypothesize a cell-oriented abstraction for developing "Virtual Tissues" with the following
assumptions: (a) tissues can be represented as a complex cellular system, (b) cells are the unit
of function and can be modeled as autonomous agents that use molecular circuits to make
decisions, and (c) injury is a collective response of the multiagent system to persistent stress.
Relevance: Current approaches for assessing the risk of adverse effects in humans are based
on animal testing, which is time consuming, resource intensive, and fraught with uncertainty.
Novel strategies are necessary to efficiently and effectively evaluate the risk of thousands of
environmental chemicals. Integrative computational systems, in vitro models, and assays offer
an avenue for more cost-effective and humane alternatives for the future of toxicity testing. Liver
toxicity is a frequent outcome in rodent testing, and it is difficult to evaluate its relevance in
humans.
Purpose: The v-Liver™ will provide in silico decision support tools to analyze the MOA in light
of available data and prior knowledge, and quantitatively simulate an MOA at environmentally
relevant tissue doses. This will inform/evolve biologically based dose-response models to
include more relevant physiologic details necessary for predicting the human risk of injury at low
doses.
Objective: The primary objective is to develop an integrated in silico/in vitro framework that aids
intelligent hypothesis generation about the plausible sequence of molecular, cellular, and tissue
events perturbed by a test chemical, and quantitatively simulates the risk of these events in
humans at environmentally relevant tissue doses using in vitro data. The v-Liver™ PoC will
focus on a subset of 20 environmental chemicals with known rodent toxicity in ToxRefDB and in
vitro data in ToxCast™. There are two specific goals of the PoC:, which are described below.
I. The v-Liver™ Knowledgebase
Knowledgebased, or semantic, approaches (Karp, 2001) are important for computationally
modeling incomplete and evolving insight on complex processes. They enable integration of
disparate biological information from literature, omic data, or pathway databases at different
scales into coherent computable representation that is flexible, extensible, and transparent.
A more important advantage of semantic approaches is their support for automated
reasoning, which is important for inferring plausible sequences of events perturbed by new
chemicals. Large-scale knowledgebased approaches have become feasible because of
semantic Web technology and available ontologies for different levels of biological
organization. The v-Liver™ Knowledgebase (v-Liver-KB) will represent normal hepatic
functions and their perturbation by chemical stressors into pathophysiologic states using
description logic expressed in OWL and stored in the Sesame semantic repository. To
facilitate the construction of the v-Liver-KB, a Cytoscape plug in is being developed to
synthesize information from different biological databases into OWL using a custom
ontology. This system also supports SPARQL-based queries, interactive visualization, and
information export in RDF. Information about the 20 PoC chemicals will be represented in
the KB at multiple biological levels describing events and causal relationships between
these events based on evidence from experiments or the literature. The main outputs of the
KB will be
(1) a computable logical description of the molecular circuits and cellular states involved in
normal hepotocyte and Kupffer cell function based on literature;
(2) a computable logical description of perturbations in molecular circuits and cell states due
to by test chemicals;
(3) interactive web-based tools to browse and interactively query/explore the v-Liver-KB to
analyze alternative MOA in light of HTS, omic or other cell based assays; and
62
-------
EPA CompTox Research Program FY2009 to 2012
Final
(4) intelligent inference tools to explore alternative pathways perturbed by chemicals based
on existing information on partial orders in the KB.
II. The v-Liver™ Simulator (Sim)
The spatial model of the hepatic lobule will be developed using a multi-agent system (MAS)
(Axelrod, 1997; Epstein and Axtell, 1996; Athale et al., 2005) in which hepatocytes and
Kupffer cells will be modeled as autonomous agents. Information on molecular circuits in the
KB will be used to describe chemical-induced molecular perturbations of NRs namely, CAR,
PXR, and PPAR-a. We are developing a variation of Probabilistic Boolean Networks
(Kauffman, 1993; Shmulevich et al. 2002) to describe the dynamics of individual agent
decisions regarding state of stress, injury, cell cycle progression, apoptosis, necrosis, or
migration (KC). These will be augmented and calibrated using available literature and/or
ToxCast™ data on the PoC chemicals. We believe this work will advance the two-stage
clonal growth models of cancer (Conolly, and Andersen, 1997) by including relevant
information on molecular pathways and cell communication. The agent population initially
will be situated in a 2-dimensional regular spatial grid to model in vitro conditions and a
simplified cross-section through the hepatic lobule. Portal-to-centrilobular blood flow initially
will be represented as a gradient of nutrients and xenobiotics (estimated from organ dose),
which can be extended to model more complex flows if necessary. Simulating the MAS will
generate a spatial distribution of cellular alterations that can be interpreted as tissue lesions.
Hence, the v-Liver™ Simulator (v-Liver-Sim) will dynamically simulate the key molecular and
cellular perturbations leading to adverse effects in hepatic tissues. The predictions will be
evaluated for the PoC chemicals using ToxCast™ data. The main outputs of the v-Liver™
Simulator are
• a large-scale tissue simulation engine to enable quantitative exploration of alternative
physiologic processes and their histopathologic outcomes.
• a computational interface to integrate the tissue simulator with a PBPK modeling system
to investigate individual exposure / population variability.
• interactive tools to communicate with the simulation engine and to visualize results of
simulations across chemicals, MOAs, doses and species.
Impact: The v-Liver™ will impact the future of toxicity testing by providing computational tools
to explore the MOA for new environmental chemicals using background knowledge, chemical
structure, and/or in vitro assays and to provide an initial assessment of hepatic lesion formation.
Focusing on 20 NR activating environmental chemicals and their hepatic lesions through a
subset of molecular pathways will demonstrate the v-Liver ™ PoC. The project also is expected
to contribute to ongoing assessments of pesticides and persistent toxics by providing useful
information about the human relevance of any liver effects and their putative dose-response. If
successful, the Virtual Tissues will be able to leverage available screening data from ToxCast™;
fill any data gaps with targeted studies; and reduce the time, the cost, and the requirement for
as many animal studies.
Synopsis: The v-Liver™ computational paradigm represents tissues as cellular systems in
which discrete individual cell-level responses give rise to complex physiologic outcomes. In this
model, cell-level responses are governed by a self-regulating network of normal molecular
processes, and adverse histopathologic effects arise because of chronic stimulation by
environmental chemicals. The v-Liver™ PoC is being developed by focusing on environmental
chemicals responsible for hepatocarcinogenesis in rodent studies, organizing MOA knowledge
on the relevant molecular and cellular processes perturbed by these chemicals, developing a
tissue simulation platform to investigate the uncertainties in MOA and neoplastic lesion
63
-------
EPA CompTox Research Program FY2009 to 2012
Final
formation, and evaluating in vitro assays to predict lesion development across chemicals and
doses.
Partnerships/Collaborations (Internal & External): EPA Collaborations: NCCT ToxCast™ in
vitro assays and their linkage with physiologic outcomes, NCCT v-Embryo™ and cell-level
modeling of tissue responses, NCCT/NHEERL/PBPK modeling to infer internal dose, NHEERL
Genomics Core on MOA for PPAR-a activators, and NHEERL PK Branch on hepatic xenobiotic
and T3/T4 metabolism. External Collaborators: UNC Carolina Center for Computational
Toxicology; UMDNJ PBPK Modeling.
Milestones/Products:
FY09
(1) Prioritize PoC environmental chemicals with clients
(2) KB: Information about PoC chemicals using ToxCast assays
(3) KB: Cytoscape KB visualization and analysis tool
(4) Cell Response: Initial molecular circuit describing hepatic cell functions
(5) Tissue Simulator: Develop and use MAS framework
FY10
(1) Tissue Simulator: Test liver lesions formation
(2) Integrate molecular circuits for MOA chemicals in tissues
(3) Evaluate simulator using PoC chemicals and ToxCast™ data to predict outcomes
FY11
(1) KB inference tool for analyzing MOA for new chemicals and mixtures
(2) Extend lobule simulator to liver and integrate with PBPK model
FY12
(1) Evaluate impact of genomic variation on cellular responses and lesion formation
(2) Evaluate v-Liver™ for simulating human pathology outcomes using clinical data
Keywords: virtual tissues, knowledgebases, MOA, dose-response modeling, nuclear receptors
mediated hepatocarcinogenesis
QA Project Plan: Category III. QA of modeling projects serves at least two overlapping goals:
(1) Verification reproducibility of results is essential for the scientific method, and (2) continuity -
proper documentation of results allows future researchers (or the same researcher after a long
period of time) to return to a project without excessive amounts of time spent understanding
what was done before. Because the v-Liver™ project requires multiple researchers working over
several years, ensuring both continuity of modeling efforts and reproducibility of modeling
results is vital. All QA plans are archived in the EPA internal QA system.
i. Uncertainty—Analysis in Toxicological Modeling
Lead/Principal Investigator: R. Woodrow Setzer
Research Issue/Relevance: The analysis of uncertainty in toxicological modeling is critical to
EPA because the Agency is increasing its use of toxicological models in regulatory decisions,
and any use of model predictions in a rational decision process must consider the uncertainty of
those predictions. The recent National Academy report on risk assessment activities in EPA
emphasizes the importance of incorporating a quantification of uncertainty in risk assessments.
In the context of models, the easiest form of uncertainty to address is that about parameter
values, assuming we have the correct model. This is the sort of uncertainty that statistical
64
-------
EPA CompTox Research Program FY2009 to 2012
Final
methodologies were designed to estimate and is typically quantified through confidence
intervals or probability distributions. However, we are rarely completely confident about the
models we use, and often, the uncertainty about the underlying processes taking place or the
best way to characterize those processes in a model can be quite substantial. This form of
uncertainty is prevalent throughout dose-response analysis, from simple empirical modeling
used in benchmark dose analysis, to pathway modeling used in virtual tissues, and is called
model uncertainty.
Both forms of uncertainty are quantified through comparisons of models with data, by
quantifying the degree to which different parameter values for a given model, and the best-fitting
parameter values among different models, yield model predictions that are consistent with the
data. "Consistency" is quantified through the mediation of an additional statistical model that
relates the biological model to the data by describing the variability in the data as it is affected
by the biological model.
In principle, this process is straightforward, and there are standard statistical procedures for
carrying out the process. However, toxicological models comprise a wide array of modeling
techniques, for example, simple algebraic expressions and systems of ordinary or partial
differential or difference equations, and may involve agent-based or stochastic process models.
Such models can be quite complex, with many parameters whose values are known with
varying degrees of uncertainty. Statistical models usually need to accommodate multiple
hierarchical levels of variability: variation among studies and among individuals, in addition to
the usual measurement error, which should be allowed to differ among studies and end points.
Information for estimating parameters and evaluating models themselves may come from well-
characterized experimental data, as well as from tabulated (for example, physiological
parameters like organ weights) or computed (for example, computed partition coefficients for a
PBPK model).
Thus, despite the existence of sound statistical theory, the application of good statistical practice
for these models can be difficult, requiring thoughtful application of both statistical and
computational expertise and quite a bit of "art" to get good results. The key challenges in
uncertainty analysis for toxicological models in risk assessment are to develop computationally
efficient, transparent, and statistically valid approaches that may be implemented without the
development of extensive case-specific programming.
Purpose/Objective/Impact: The objective of this project is to develop tools and best practices
to facilitate the quantification of uncertainty in toxicological models. Early efforts will focus on
PBPK models, specifically models being developed as part of a joint project between the
Agency's ORD and OPP to develop a cumulative risk assessment for the pyrethroid pesticides.
However, similar methods are applicable to other forms of toxicological models, and examples
will be pursued in the virtual tissue projects in NCCT and in evaluating uncertainty in the
ToxCast™ predictors.
In recent years, it has become clear that Bayesian statistical methods are ideal for estimating
parameters for PBPK models. For three reasons: (1) the ease with which partial information
about parameters is included, through informative priors; (2) the ease with which hierarchical
models are constructed; and (3) the fact that, as long as informative priors are used, lack of
identifiability of model parameters, which is not generally possible to diagnose a priori, is not an
impediment to completing a valid analysis (failure of identifiability of parameters may be
diagnosed from analyses of the posterior parameter distribution).
65
-------
EPA CompTox Research Program FY2009 to 2012
Final
The same logic should apply to other toxicological models. Thus, this project will center on tools
to apply Bayesian methods to such models.
More specific objectives of this project are those that follow.
• Develop the computational tools to carry out Bayesian analyses of large dynamic models as
efficiently as possible. The standard approach to evaluating the posterior in a Bayesian
analysis of complex models is to generate random samples from the distribution by Markov
Chain Monte Carlo (MCMC) methods. For models that are expensive to compute, such as
those (e.g., PBPK models) expressed as solutions to systems of differential equations, an
implementation that takes advantage of cluster computing may take advantage of some
parallel structures in the problem. This objective will develop a standard computational
approach to parallelizing such problems and will explore alternative approaches to
implementing MCMC methods, with an eye towards computational efficiency. This objective
also includes the development of modeling tools to facilitate dynamic models in the statistical
language R.
• Describe a language specifically for expressing PBPK models. The language should be
extensible, be capable of incorporating systems models expressed in SBML, and should use
semantics specific to PBPK models to facilitate model checking. The language would be ideal
as a way to archive PBPK models and as the definitive way to communicate PBPK models in
the literature.
• Adapt statistical model evaluation approaches to complex toxicological models, and develop
examples to demonstrate the behavior of different model evaluation methodologies in the face
of various model failures.
• Develop a general approach to developing priors for chemical-specific PBPK model
parameters. There are already methods for computing chemical-specific parameters either
from physical chemical properties or in vitro assays (depending on the parameter). For this
objective, datasets of measured chemical-specific parameters will be compared to values
predicted from computational or in vitro methods. Statistical methods, such as regression will
be used to adjust the predictions and the variance about the resulting regression, lines used
to characterize the prior uncertainty about such predictions.
• Develop approaches for quantifying uncertainty of ToxCast™ like predictors involving HTS
data as inputs, explicitly evaluating the importance of variability and design of HTS assays on
prediction and prioritization uncertainty.
• Identify examples for parameter estimation, model evaluation, and overall uncertainty analysis
drawn from PBPK models for pyrethroid pesticides and the encompassing cumulative risk
analysis being conducted in collaboration with NHEERL, NERL, and OPP; selecting among
molecular and cellular models used in developing a virtual liver model, and others developed
in collaboration with the virtual tissues projects in NCCT.
Synopsis: This plan will target three areas: (1) standardizing and making more efficient
computational approaches for parameter estimation and model selection; (2) standardizing
approaches for model evaluation for PBPK and other dynamic models; and (3) development of
methods for constructing priors (probabilistic summaries of current knowledge) for model
parameters, based on existing computational methods and datasets. The initial motivation for
this work was the need to standardize and make more sophisticated parameter estimation and
model evaluation for PBPK models, and that emphasis will continue in the early phase of this
project. However, all models relevant to toxicological risk assessment have similar
requirements, and this project will coordinate closely with the virtual tissue and dose-response
projects.
66
-------
EPA CompTox Research Program FY2009 to 2012
Final
Partnerships/Collaborations (Internal & External): Internal: Jimena Davis (postdoctoral
Fellow); Richard Judson, John Wambaugh, Imran Shah, and Thomas Knudsen.
External: ORD/NHEERL: Mike Hughes, Kevin Crofton, Tim Shafer, Ginger Moser, Rory Conolly;
ORD/NERL: Rogelio Tornero, Valerie Zartarian, Xianping Xue; OPPTS/OPP: Anna Lowit, David
Miller, Ed Scollon; and NIEHS/NTP: Mike DeVito.
Milestones/Products:
FY09
(1) Contribution to 2009 SOT CED course, "Uncertainty and Variability in PBPK Models"
(2) Submission of manuscript(s) on improvement of computational efficiency in Bayesian
analyses of PBPK models using MCMC, and assessment of convergence (Wambaugh,
Davis, Garcia, and Setzer)
(3) Submission of manuscript(s) on assessing fit of PBPK models (and, through example, other
complex mechanistic models) to data, whether the models are "fit" to the data using
Bayesian or other methods are parameterized a priori from in vitro data (Wambaugh, Davis,
Garcia, and Setzer)
(4) Submission of manuscript(s) reviewing approaches for constructing priors for PBPK model
parameters (i.e., establishing a priori estimates for model parameters in advance of using in
vivo PK data, with a characterization of uncertainty) (Davis, Setzer, Tornero, and DeVito)
(5) Draft manuscript(s) on the estimation of model parameters and comparison of alternative
model forms for a PBPK model for permethrin (in preparation of SAP review in FY10)
(Davis, Setzer, Tornero, and DeVito)
(6) Draft manuscript(s) on global sensitivity analysis for exposure-dose model for permethrin, in
preparation for SAP review in FY10. (Davis, Setzer, Tornero, Zartarian, and Chiu)
(7) Estimation of PBPK parameters for deltamethrin and two other pyrethroids complete Davis,
Setzer, and Tornero)
FY10
(1) Manuscript on permethrin PBPK parameter estimation and model comparison submitted
(2) Manuscript on global sensitivity analysis for permethrin exposure-dose model submitted
(3) SAP review of permethrin PBPK exposure-dose model
(4) Submission of manuscript on parameter estimation for deltamethrin and two other
pyrethroids
(5) Completion of exposure-dose-effect model for "mini-cumulative" risk assessment and draft
document for SAP review in early FY11, including global "exposure to effect" sensitivity
analysis (Davis, Setzer, Tornero, DeVito, Crofton, Shafer, Lowit, Scollon, Miller, etc.)
(6) Substantial completion of modeling and uncertainty analysis for v-Liver™ (Setzer, Shah, and
Wambaugh)
(7) R package "RDynamic", for simplifying dynamic modeling in the statistical language R,
submitted to CRAN
(8) Description of ontologically aware PBPK language working name "SemanticPK" drafted, and
translator to R completed (using code developed for RDynamic)
(9) Discussion of uncertainty about ToxCast™ phase I predictions at 2010 SOT.
(10) Problem identification and early stages of evaluating alternative pathway formulations in
v-Liver™ model (Setzer, Shah, and Wambaugh).
FY11
(1) Problem identification, timelines, and early stages of evaluating alternative model
formulation for BBDR and v-embryo™ models completed (Setzer, Conolly, and Knudsen
(2) SAP on "mini-cumulative" risk assessment for pyrethroids
(3) Preparation for SAP for cumulative risk assessment for pyrethroids
(4) Publication of virtual tissue and BBDR modeling results
(5) Initial steps in constructing test datasets for analysis "competition" recommended by UVPKM
67
-------
EPA CompTox Research Program FY2009 to 2012
Final
(6) Submission of manuscript on "RDynamic" to the Journal of Statistical Software
FY12
(1) Submission of manuscript on "SemanticPK"
(2) R package "RDynamic", for simplifying dynamic modeling in the statistical language R,
submitted to CRAN
Keywords: uncertainty analysis, physiologically based pharmacokinetic models, statistics,
Bayes methods, priori, Markov Chain Monte Carlo
QA Project Plan: Category III. This project is a Category III QA category because of the
significance of the pyrethroid cumulative risk assessment to which this plan contributes. Several
quality objectives apply to this project (1) Computer code developed for the project must
faithfully execute the intent of the code, whether that intent is described mathematically or in
terms of other software (for example, implementations of the same model in different
programming languages must give identical results for identical inputs); (2) distribution versions
of software packages must be installable and useable by a reasonably sophisticated person; (3)
analyses must be transparent: It must be possible to replicate all analyses from the archived
files and information; and (4) data sources must be transparent; in particular, data from the
literature must be annotated to be adequate for purpose, and extent of literature searches must
be documented. All QA plans are archived in the EPA internal QA system.
68
-------
EPA CompTox Research Program FY2009 to 2012 Final
IV. APPENDICES (cont'd.) - 2. Project Outcomes
2. Project Outcomes - Providing High-Throughput Computational Tools for the Identification of Chemical Exposure, Hazard, and Risk
Project Title
Outcomes FY09
Outcomes FY10
Outcomes FY11
Outcomes FY12
Expected Impacts
ACToR - Aggregated
Computational
Toxicology Resource
• Initial public deployment.
• Significant version 2,
including refined chemical
structure information
• Develop workflow for
tabularization of data buried
in text reports
• Integrate all ToxCast and
ToxRefDB data
• Quarterly releases with new
data
• Quarterly releases with new
data
• Implementation of a process
to gather tabular data on
priority chemicals from text
reports
• Survey sources of chemical
use and exposure data and
import any remaining sources
• Develop flexible query
interface and data download
process
• Develop process to extract
data from open literature
• Quarterly releases with new
data
• Quarterly releases with new
data
Enables access to cross-chemical data by EPA
program offices, NCCT, and other ORD organizations
and external stakeholders. Improves EPA data
transparency.
DSSTox- Chemical
Information
Technologies in
Support of Toxicology
Modeling
• MS: SAR Perspective of
ToxCast 320 chemical
inventory
• MS(s): NCBI (GEO) and
EBI (ArrayExpress) structure
annotations and linkages to
micro array data
• Restart Chemoinformatics
CoP
• Publish files for ToxReDB
and ToxCast inventories and
selected summary end
points, and facilitate
publication and linkage to
PubChem
• Publish public genetic
toxicity data and SAR
predictions for ToxCast 320
• Continue expansion of
DSSTox public toxicity
database inventory.
• Primary chemical review
and structure annotation of
ToxCast/Tox21 libraries
within a central registry
• Publish files for Tox21
inventory and selected
summary end points and
facilitate linkage to PubChem
• Publish files for NTP study
areas
• Explore new approaches to
SAR based on feature
categories
• Expand CEBS collaboration
to incorporate DSSTox GEO
and ArrayExpress files and
create chemical linkage to
ILSI Developmental Toxicity
database
• Assist efforts within
ExpoCast regarding
chemically annotation
• Establish procedures and
protocols for automating
chemical annotation of new
experimental data generated
by NCCT and in collaboration
with CEBS or NHEERL
• Document and employ
PubChem analysis tools in
relation to published DSSTox
and ToxCast data
• Collaborate with SAR
modeling efforts to predict
ToxCast end points
• Continue expansion of
DSSTox public toxicity
database inventory for use in
modeling with co-publication
and linkage to ACToR and
PubChem
• Redesign DSSTox Website
to provide hosting of donated
chemical descriptors,
properties, and predictions
• Publish master tables of
DSSTox IDs and high quality
structures
• Promote use of chemical
registry system from
ToxCast/Tox21 more broadly
within EPA
• Collaborate with SAR
modeling efforts to expand
modeling to address Tox21
chemicals and endpoints
• Continue expansion of
DSSTox public toxicity
database inventory into new
toxicity and exposure areas
Adoption of DSSTox chemical standards more
broadly within EPA and across other government
Agencies to improve quality and read-across
capabilities.
Promote public data dissemination and encourage
greater public participation of industry and
commercial sources in public toxicity database and
modeling efforts.
Facilitate improved toxicity prediction models and
data mining capabilities across wider span of end
points and chemicals impacting hazard ID and risk
assessment.
69
-------
EPA CompTox Research Program FY2009 to 2012 Final
IV. APPENDICES (cont'd.) - 2. Project Outcomes
2. Project Outcomes - Providing High-Throughput Computational Tools for the Identification of Chemical Exposure, Hazard, and Risk
Project Title
Outcomes FY09
Outcomes FY10
Outcomes FY11
Outcomes FY12
Expected Impacts
ToxRefDB-Toxicity
Reference Database
• MS(s): Chronic/cancer,
multigenerational and
developmental modules
• Release of stand-alone data
entry tool
• ToxRefDB Webpage online.
• Collection of ToxCast
Phase II chemical toxicity
data
• Public release of ToxRefDB
Web-based query tool
• Complete entry of targeted
set of chemicals and study
types for Phase II of ToxCast
• Complete reproductive
toxicity study retrospective
analysis
• Quarterly releases with new
data in conjunction with
ACToR (Q1,Q2,Q3,Q4)
• Implementation of a process
to gather and enter open
literature studies (Q1)
• Expansion of ToxRefDB to
capture DNT studies and
EDSPdata (Q1)
• Complete retrospective
analyses on other major
study types (Q2)
• Release of ToxRefDB live
data entry tool
• Quarterly releases with new
data in conjunction with
ACToR
• Quarterly releases with new
data in conjunction with
ACToR
Enables access of traditional toxicological data in a
structured and computable format extending the utility
of the data beyond chemical risk assessment and into
broad research applications, including guiding novel
toxicity characterization methods and transparent
data-driven retrospective analyses leading to refined
animal use, a more predictive toxicology paradigm,
and more efficient chemical safety assessments.
ChemModel -The
Application of Molecular
Modeling to Assessing
Chemical Toxicity
• MS: Capability of the target-
toxicant paradigm to identify
chemicals that binds weakly
to the estrogen receptor,
including a description of the
method
• Description of the library of
148 biological
macromolecule targets
• MS: Molecular modeling
studies of the potential
biological effects of the
perfluoro compounds
• MS: The metabolism of
pyrethroids and the effects of
three-dimensional chemical
structure
• Description of additional
targets added to the target
library
• MS: The interaction of
ToxCast chemicals with
nuclear receptor targets and
the importance of
pharmacophore filters
• MS: The integration of
results from the target library
and available experimental
parameters.
• MS: The comparison of
results with and without
pharmacophore filters for
ToxCast chemicals
• Comparison of results using
the target library with
experimental determined
activities
• MS: The use of molecular
modeling and the target
toxicant paradigm for
regulatory purposes, including
a discussion of the OECD
principles relative to molecular
modeling specific principles
An approach will be provided to prioritize chemicals
for their ability to influence the endocrine system by
competing with natural ligands for the binding sites of
receptors. The application of a method used to find
strong-acting drug-like chemicals will be evaluated for
its capability of discovering weakly active chemicals.
The capability of these methods will be used for
finding weakly active chemicals. Parameters based
on the capacity of a chemical to interact with
macromolecular targets for toxicity will be available
for applications of computational methods to screen
for or, eventually, predict chemical toxicity.
70
-------
EPA CompTox Research Program FY2009 to 2012 Final
IV. APPENDICES (cont'd.) - 2. Project Outcomes
2. Project Outcomes - Providing High-Throughput Computational Tools for the Identification of Chemical Exposure, Hazard, and Risk
Project Title
Outcomes FY09
Outcomes FY10
Outcomes FY11
Outcomes FY12
Expected Impacts
ToxCast™-Screening
and Prioritization of
Environmental
Chemicals Based on
Bioactivity Profiling and
Predictions of Toxicity
• Completion of ToxCast
Phase 1
• Provide Phase 1 data sets to
public
• Derivation of predictive
signatures from ToxCast
Phase 1 data
• MS(s): ToxCast Phase 1
data sets
• MS: Signature generation
• MS: NR pathways and
toxicity
• Convene first ToxCast Data
Summit for identifying
prediction models
• Finalize selection of
chemicals for Phase II of
ToxCast and Tox21
• MS(s): Describing
approaches to combining
exposure, PK, and in vitro
assays to do risk prioritization
• Evaluate compatibility of
nanomaterials of diverse
classes with ToxCast assays
• Prioritize and select assays
to be run for ToxCast Phase
II
• Completion of ToxCast
Phase II data collection
• Confirmation of ToxCast
predictive signatures with
Phase II data
• MS(s): Signature
confirmations and
applications
• MS(s): Profiling of large
chemical sets for potential for
hazard
Program offices will have tools to prioritize chemicals
for targeted toxicity testing and insight on potential
mechanisms of toxicity.
ExpoCast™: Exposure
Science for Screening,
Prioritization, and
Toxicity Testing
• ExpoCoP monthly
teleconference, ESC
resource, face-to-face
meeting at ISES 2009
• MS(s): EHP and ToxSci on
role of exposure in the
transforming toxicology
• SVOC workshop, "Semi-
Volatile Organic Compounds
(SVOCs) in the Residential
Environment"
• Survey and identify high
priority exposure data
resources for initial chemical
indexing in collaboration with
ACToR and DSSTox
• ExpoCast conceptual
framework and research plan
• Position paper
recommending standards for
exposure data representation
• White paper defining
exposure space and plan for
assessing exposure data
landscape
• White paper exploring
development and application
of human exposure
knowledge base
• Begin implementation of
standards across exposure
databases of highest interest
and utility for NCCT projects
• Award contracts
• MS: Describing extant data
analyses to identify critical
determinants for exposure
classification and chemical
prioritization based on
potential for exposure
• Guide incorporation and
further development of simple
exposure estimation tools
within the ACToR system for
use in prioritization
• Apply exposure index for
prioritization to subset of
ToxCast compounds to
evaluate concept
Advance Agency tools for efficiently characterizing
and classifying chemicals based on potential for
biologically relevant exposure and inform
characterization of environmentally relevant toxicity.
71
-------
EPA CompTox Research Program FY2009 to 2012 Final
IV. APPENDICES (cont'd.) - 2. Project Outcomes
2. Project Outcomes - Providing High-Throughput Computational Tools for the Identification of Chemical Exposure, Hazard, and Risk
Project Title
Outcomes FY09
Outcomes FY10
Outcomes FY11
Outcomes FY12
Expected Impacts
v-Embryo™-Virtual
Embryo
• MS: Application of VT-KB to
analyze ToxRefDB
developmental toxicity
studies
• VT-KB based qualitative
(structural) model of self-
regulating ocular gene
network
• VT-SE based cell-based
computational model of lens-
retina induction
• MS: Ocular morphogenesis,
gene network inference,
analysis and modeling
• Extend lens-retina model to
other stages and species
• Incorporate pathway data
from ToxCast, mESC and ZF
embryos
• MS: Sensitivity analysis for
key biological pathways
• MS: Developmental
trajectories and phenotypes
in computational models
• Integrate with other
morphogenetic models (ES
cells, Zfish)
• MS: Test of model against
predictions for pathway-
based dose-response
relationship
• MS: Uncertainty analysis of
models for complex systems
model
• Model: Computer program
of early eye development
using rules-based
architecture, cell-based
simulators, and systems-
wiring diagrams
• MS: Integrated eye
morphogenesis
• Integration of computational
models of different systems
• Evaluation of integrative
model with data from ES cells,
Zfish
Framework for in silico reconstruction of the embryo
to facilitate navigation of complex relationships and
predict systems-level behavior (outcome) from data
on biochemical, molecular, and cellular changes.
v-Liver™ - The Virtual
Liver Project
• Prioritize proof of concept
(PoC) environmental
chemicals with clients.
• Knowledge Base (KB):
Information about PoC
chemicals using ToxCast
assays
• KB: Cytoscape KB
visualization and analysis
tool
• Cell response: Initial
molecular circuit describing
hepatic cell functions
• Tissue Simulator: Develop /
use MAS framework
• Tissue Simulator: Test liver
lesions formation
• Integrate molecular circuits
for MOA chemicals in Tissues
• Evaluate simulator using
PoC chemicals and ToxCast
data to predict outcomes
• KB inference tool for
analyzing MOA for new
chemicals/mixtures
• Extend lobule simulator to
liver and integrate with PBPK
model
• Evaluate impact of genomic
variation on cellular responses
and lesion formation
• Evaluate v-Liver for
simulating human pathology
outcomes using clinical data
Proof-of-concept decision support tools to enable
tiered-testing:-
(1) reduce uncertainty in evaluating the effect of
chemicals on normal hepatic pathways and
(2) estimate dose-dependent adverse hepatic effects
in individuals and variability in populations.
Uncertainty - Analysis in
Toxicological Modeling
•MS: Improvement of
computational efficiency in
Bayesian analyses of PBPK
models using MCMC
• MS: Assessing fit of PBPK
models to data using
Bayesian or other methods
• MS: Reviewing approaches
for constructing priors for
PBPK model parameters
• MS: Estimation of model
parameters and comparison
of alternative model forms for
permethin PBPK model
• MS: Global sensitivity
analysis for exposure-dose
model for permethrin
• Estimation of PBPK
parameters for deltamethrin
and two other pyrethroids
complete
• MS:Permethrin PBPK
• MS on global sensitivity
analysis for permethrin
exposure-dose model
submitted
• MS: Parameter estimation
for deltamethrin and two
other pyrethroids
• Completion of exposure-
dose-effect model for "mini-
cumulative" risk assessment
• Substantial completion of
uncertainty analysis for v-
Liver
• Description of ontologically
aware PBPK language and
translation to R
• Problem identification for
evaluating alternative
pathway formulations in v-
Liver model
• Problem identification
evaluating alternative model
formulation for BBDR and
virtual embryo models
• SAP on "mini-cumulative"
risk assessment for
pyrethroids
• Preparation for SAP for
cumulative risk assessment
for pyrethroids
• Publication of virtual tissue
and BBDR modeling results
• Initial steps in constructing
test datasets for analysis
"competition" recommended
by UVPKM
• Submission of ms on
"RDynamic" to the Journal of
Statistical Software
• Submission of MS on
"Semantic-PK"
• R package "RDynamic", for
simplifying dynamic modeling
in the statistical language R,
submitted to CRAN
Improved cumulative risk assessment for pyrethroid
pesticides, based on realistic quantitative assessment
of uncertainties.
72
-------
EPA CompTox Research Program FY2009 to 2012
Final
IV. APPENDICES (cont'd.)
B. Extramural STAR Centers Projects
1. The Research Center for Environmental Bioinformatics and Computational Toxicology at the
University of Medicine and Dentistry of New Jersey (UMDNJ), Piscataway, brings together a
team of computational scientists with diverse backgrounds in bioinformatics, chemistry, and
environmental science, from UMDNJ, Rutgers, and Princeton Universities, and the U.S. Food
and Drug Administration's Center for Toxicoinformatics. The team is addressing multiple
elements of the source-to-outcome sequence for toxic pollutants, as well as developing tools for
toxicant characterization. The computational tools developed through this effort will be
evaluated extensively and refined through collaboration between STAR Center scientists and
with colleagues from the three universities and EPA. Particular emphasis is on methods that
enhance current risk assessment practices and reduce uncertainties. Researchers also are
developing the Web-accessible Environmental Bioinformatics Knowledge Base that will provide
a user-oriented interface to an extensive set of information and modeling resources.
2. The Carolina Environmental Bioinformatics Research Center at the University of North
Carolina, Chapel Hill, is developing new analytic and computational methods, creating efficient
user-friendly tools to disseminate the methods to the wider community, and applying the
computational methods to molecular toxicology and other studies. The center brings together
multiple investigators and disciplines, combining expertise in biostatisties, computational
biology, chemistry, and computer science to advance the field of computational toxicology.
Researchers focus on providing biostatistician support to the center by performing analyses and
developing new methods in computational biology. The center is also creating a framework for
merging data from various technologies in a systems-biology approach.
3. The Carolina Center for Computational Toxicology at the University of North Carolina, Chapel
Hill, will advance the field of computational toxicology through the development of new methods
and tools, as well as through collaborative efforts. The center is utilizing a bottom-up approach
to predictive computational modeling of adverse effects of toxic agents. The emphasis spans
from the fine-scale predictive simulations of the protein-protein/-chemical interactions in nuclear
receptor networks, to mapping chemical-perturbed networks and devising modeling tools that
can predict the pathobiology of the test compounds based on a limited set of biological data, to
building tools that will enable toxicologists to understand the role of genetic diversity between
individuals in responses to toxicants, to unbiased discovery-driven prediction of adverse chronic
in vivo outcomes based on statistical modeling of chemical structures, high-throughput
screening and the genetic makeup of the organism. In each project, new computer-based
models will be developed and published that represent the state of the art. The tools produced
within each project will be disseminated widely, and the emphasis will be placed on their
usability by the risk assessment community and investigative toxicologists alike. The synthesis
of data from a variety of sources will move the field of computational toxicology from a
hypothesis-driven science toward a predictive science.
4. Texas-Indiana Virtual STAR Center; Data-Generating in vitro and in silico Models of
Developmental Toxicity in Embryonic Stem Cells and Zebrafish at The University of Houston,
Texas A&M Institute for Genomic Medicine, and Indiana University. The project period is from
November 1, 2009 to October 31, 2012.
Objectives/Hypothesis:
73
-------
EPA CompTox Research Program FY2009 to 2012
Final
As chemical production increases worldwide, there is increasing evidence as to their hazardous
effects on human health at today's exposure levels, which further implies that current chemical
regulation is insufficient. Thus, a restructuring of the risk assessment procedure will be required
to protect future generations. Given the very large number of man-made chemicals and the
likely complexity of their various and synergistic modes of action, emerging technologies will be
required for the restructuring. The main objective of the proposed multidisciplinary Texas-
Indiana Virtual STAR (TIVS) Center is to contribute to a more reliable chemical risk assessment
through the development of high-throughput in vitro and in silico screening models of
developmental toxicity. Specifically, the TIVS center aims to generate in vitro models of murine
embryonic stem cells and zebrafish for developmental toxicity. The data produced from these
models will be exploited further to produce predictive in silico models for developmental toxicity
on processes that are relevant also for human embryonic development.
Approach:
The project is divided into three investigational areas; zebrafish models, murine embryonic stem
cells models, and in silico simulations. The approaches are as follows.
(1) Generate developmental models suitable for high-throughput screening. Zebrafish
developmental models (transgenic GFP/EGFP/RFP models of crucial steps in development)
and embryonic stem cell (ESC) differentiation models (transgenic beta-geo models of crucial
steps in differentiation) will be generated. Important morphology features and signaling
pathways during development will be documented. The impact of environmental pollutants
on development and differentiation will be assessed in the models. Finally, the models will
be refined for high-throughput screening and automation.
(2) Generate a computational model that faithfully recreates the major morphological features of
normal wild-type zebrafish development (i.e., segmentation into somites, proper patterning
of vascular and neural systems) and the differentiation to three primitive layers (endoderm,
mesoderm, and ectoderm) in mouse embryonic stem cells. The data for simulations are
produced from developed high-information-content zebrafish and ESC models. Once a
working model of normal development has been generated, we will carry out a directed
series of parameter sweeps to try to create developmental defects in silico. We will compare
the results of computationally created defects with experimentally generated defects in
zebrafish and embryonic stem cells. Best matches between the two datasets will suggest
hypotheses about possible mechanisms by which defects occur.
(3) Perform proof-of-concept experiments of the in vitro and in silico test platforms with a blind
test of chemicals.
Techniques will be molecular biology techniques on zebrafish and ESC models, such as
cloning, imaging, in vitro differentiation and in vitro exposure studies, and in silico mathematical
simulations.
Expected Results (Outputs/Outcomes):
In collaboration with other initiatives taken in the field of chemical safety, our generated results
and models will contribute to large screening effort to prioritize chemicals for further risk
assessment. We will contribute specifically with
• 9 transgenic fish lines validated for toxicity screening;
• 16 embryonic stem cell models validated for toxicity screening;
• High-information-content models on development and differentiation to produce data for in
silico simulations, within the project and elsewhere;
• computational models for developmental toxicology of normal development and of
mechanisms by which chemical perturbations cause experimentally observed developmental
defects; and
• Information on developmental toxicity on 39 compounds.
74
-------
EPA CompTox Research Program FY2009 to 2012
Final
C. FY2004 "New Start" Award Bibliography
Project Title: Linkage of Exposure and Effects Using Genomics, Proteomics, and
Metabolomics in Small Fish Models
Peer-Reviewed Publications
Ankley, G.T., K.M. Jensen, E.J. Durhan, E.A. Makynen, B.C. Butterworth, M.D. Kahl, D.L.
Villeneuve, A. Linnum, L.E. Gray, M. Cardon and V.S. Wilson. 2005. Effects of two fungicides
with multiple modes of action on reproductive endocrine function in the fathead minnow
(Pimephales promelas). Toxicol. Sci. 86, 300-308.
Ankley, G.T., K.M. Jensen, M.D. Kahl, E.A. Makynen, L.S. Blake, K.J. Greene, R.D. Johnson
and D.L. Villeneuve. 2007. Ketoconazole in the fathead minnow (Pimephlaes promelas):
reproductive toxicity and biological compensation. Environ. Toxicol. Chem. 26, 1214-1223.
Ankley, G.T., D.H. Miller, K.M. Jensen, D.L. Villeneuve and D. Martinovic. 2008.
Relationship of plasma sex steroid concentrations in female fathead minnows to reproductive
success and population status. Aquat. Toxicol. 88, 69-74.
Ankley, G.T., D. Bencic, M. Breen, T.W. Collette, R. Connolly, N.D. Denslow, S. Edwards, D.R.
Ekman, K.M. Jensen, J. Lazorchak, D. Martinovic, D.H. Miller, E.J. Perkins, E.F. Orlando, N.
Garcia-Reyero, D.L. Villeneuve, R.-L.Wang and K. Watanabe. 2009. Endocrine disrupting
chemicals in fish: Developing exposure indicators and predictive models of effects based on
mechanisms of action. Aquat. Toxicol. 92, 168-178.
Breen, M.S., D.L. Villeneuve, M. Breen, G.T. Ankley and R.B. Conolly. 2007. Mechanistic
computational model of ovarian steroidogenesis to predict biochemical responses to endocrine
active compounds. Ann. Biomed. Engin. 35, 970-981.
Ekman, D.R., Q. Teng, K.M. Jensen, D. Martinovic, D.L. Villeneuve, G.T. Ankley and T. W.
Collette. 2007. NMR analysis of fathead minnow urinary metabolites: a potential approach for
studying impacts of chemical exposures. Aquat. Toxicol. 85, 104-112.
Ekman, D.R., Q. Teng, D.L. Villeneuve, M.D. Kahl, K.M. Jensen, E.J. Durhan, G.T. Ankley and
T.W. Collette. 2008. Investigating compensation and recovery of fathead minnow (Pimephales
promelas) exposed to 17a-ethynylestradiol with metabolite profiling. Environ. Sci. Tecnhol. 42,
4188-4195.
Ekman, D.R., Q. Teng, D.L. Villeneuve, M.D. Kahl, K.M. Jensen, E.J. Durhan, G.T. Ankley and
T.W. Collette. 2009. Profiling lipid metabolites yields unique information on gender-and time-
dependent responses of fathead minnows (Pimephales promelas) exposed to 17a-
ethynylestradiol. Metabolomics 5, 22-32.
Garcia-Reyero, N., D.L. Villeneuve, K.J. Kroll, L. Liu, E.F. Orlando, K.H. Watanabe, M.S.
Sepulveda, G.T. Ankley and N.D. Denslow. 2009. Expression signatures for a model androgen
and antiandrogen in the fathead minnow ovary. Environ. Sci. Technol. 43, 2614-2619.
Garcia-Reyero, N., K.J. Kroll, L. Liu, E.F. Orlando, K.H. Watanabe, M.S. Sepulveda, D.L.
Villeneuve, E.J. Perkins, G.T. Ankley and N.D. Denslow. 2009. Gene expression responses in
male fathead minnows exposed to binary mixtures of an estrogen and antiestrogen. BMC
Genomics, 10:308.
75
-------
EPA CompTox Research Program FY2009 to 2012
Final
Johns, S.M., M.D. Kane, N.D. Denslow, K.H. Watanabe, E.F. Orlando, D.L. Villeneuve,
G.T. Ankley and M.S. Sepulveda. 2009. Characterization of ontogenetic changes in gene
expression in the fathead minnow (Pimephales promelas). Environ. Toxicol. Chem. 28, 873-880.
Martyniuk, C.J., S. Alvarez, S. McClung, D.L. Villeneuve, G.T. Ankley and N.D. Denslow. 2009.
Quantitative proteomic profiles of androgen receptor signaling in the liver of fathead minnows
(Pimephales promelas) J. Proteome Res. (8)5:2186-200.
Martinovic, D., L.S. Blake, E.J. Durhan, K.J. Greene, M.D. Kahl, K.M. Jensen, E.A. Makynen,
D.L. Villeneuve and G.T. Ankley. 2008. Characterization of reproductive toxicity of vinclozolin in
the fathead minnow and co-treatment with an androgen to confirm an anti-androgenic mode of
action. Environ. Toxicol. Chem. 27, 478-488.
Miller, D.H., K.M. Jensen, D.L. Villeneuve, M.D. Kahl, E.A. Makynen, E.J. Durhan and G.T.
Ankley. 2007. Linkage of biochemical responses to population-level effects: a case study with
vitellogenin in the fathead minnow (Pimephlaes promelas). Environ. Toxicol. Chem. 26, 521-
527.
Perkins, E.J., N. Garcia-Reyero, D.L. Villeneuve, D. Martinovic, S.M. Brasfield, L.S.
Blake, J.D. Brodin, N.D. Denslow and G.T. Ankley. 2008. Perturbation of gene
expression and steroidogenesis with in vitro exposure of fathead minnow ovaries to
ketoconazole. Mar. Environ. Res. 66, 113-115.
Villeneuve, D.L., P. Larkin, I. Knoebl, A.L. Miracle, M.D. Kahl, K.M. Jensen, E.A. Makynen, E.J.
Durhan, B.J. Carter, N.D. Denslow and G.T. Ankley. 2007. A graphical systems model to
facilitate hypothesis-driven ecotoxicogenomics research on the brain-pituitary-gonadal axis.
Environ. Sci. Technol. 40, 321-330.
Villeneuve, D., L. Blake, J. Brodin, K. Greene, I. Knoebl, A. Miracle, D. Martinovic and G.T.
Ankley. 2007. Transcription of key genes regulating gonadal steroidogenesis in control and
ketoconazole- or vinclozolin-exposed fathead minnows. Toxicol. Sci. 98, 395-407.
Villeneuve, D.L., L.S. Blake, J.D. Brodin, J.E. Cavallin, E.J. Durhan, K.M. Jensen, M.D. Kahl,
E.A. Makynen, D. Martinovic, N.D. Mueller and G.T. Ankley. 2008. Effects of a 3(3-
hydroxysteroid dehydrogenase inhibitor, trilostane, on the fathead minnow reproductive axis.
Toxicol. Sci. 104, 113-123.
Villeneuve, D.L., N.D. Mueller, D. Martinovic, E.A. Makynen, M.D. Kahl, K.M. Jensen, E.J.
Durhan, J.E. Cavallin, D. Bencicand G.T. Ankley. 2009. Direct effects, compensation and
recovery in female fathead minnows exposed to a model aromatase inhibitor. Environ. Health
Perspect. 117, 624-631.
Villeneuve, D.L., R.L. Wang, D.C. Bencic, A.D. Biales, D. Martinovic, J.M., Lazorchak, G. Toth
and G.T. Ankley. 2009. Altered gene expression in the brain and ovaries of zebrafish exposed
to the aromatase inhibitor fadrozole: microarray analysis and hypothesis generation. Environ.
Toxicol. Chem. (8): 1767-82.
Wang, R.L., A. Biales, D. Bencic, D. Lattier, M. Kostich, D. Villeneuve, G.T. Ankley, J.
Lazorchak and G. Toth. 2008a. DNA microarray application in ecotoxicology: experimental
design, microarray scanning, and factors impacting transcriptional profiles in a small fish
species. Environ. Toxicol. Chem. 27, 652-663.
76
-------
EPA CompTox Research Program FY2009 to 2012
Final
Wang, R.L., D. Bencic, A. Biales, D. Lattier, M. Kostich, D. Villeneuve, G.T. Ankley, J.
Lazorchak and G. Toth. 2008b. DNA microarray-based ecotoxicological discovery in a small fish
species. Environ. Toxicol. Chem. 27, 664-675.
Watanabe, K.H., K.M. Jensen, E.F. Orlando and G.T. Ankley. 2007. What is normal? A
characterization of the values and variability in reproductive endpoints of the fathead minnow,
Pimephales promelas. Comp. Biochem. Physiol. 146, 348-356.
Watanabe, K.H., Z. Li, K. Kroll, D.L. Villeneuve, N.J. Szabo, E.F. Orlando, M.S.
Sepulveda, T.W. Collette, D.R. Ekman, G.T. Ankley and N.D. Denslow. 2009. A physiologically-
based model of endocrine-mediated responses of male fathead minnows to17a-ethinylestradiol.
Toxicol. Sci. in press.
Project Title: Simulating Metabolism of Xenobiotic Chemicals as a Predictor of Toxicity
Peer-Reviewed Publications
Kenneke, J. F. 2006. Environmental fate and ecological risk assessment for the reregistration of
antimycin A (PC Code 006314), Appendix D: In vitro mammalian metabolism and Appendix G:
Summary of antimycin hydrolysis research, U.S. EPA, Office of Pesticide Programs,
Reregistration Eligibility Decision (RED) on Antimycin A.
Mazur, C. S.and Kenneke, J. F. 2008. Cross-species comparison of conazole fungicide
metabolites using rat and rainbow trout, (Onchorhynchus mykiss) hepatic microsomes and
purified human CYP 3A4. Environmental Science and Technology, 42:947-954.
Mazur, C. S.; Kenneke, J.F.; Tebes-Stevens, C. Okino, M. S.; Lipscomb, J. C. 2007. In vitro
metabolism of the fungicide and environmental contaminant trans-bromuconazole and
implications for risk assessment. Journal of Toxicology and Environmental Health, Part A, 70:
1241-1250.
Project Title: Risk Assessment of the Inflammogenic and Mutagenic Effects of Diesel
Exhaust Particulates: A Systems Biology Approach
Peer-Reviewed Publications
Cao, D., Bromberg, PA and Samet, JM. 2007. Diesel-induced Cox-2 expression involves
chromatin modification via degradation of HDAC1 and recruitment of p300. Am. J. Respir. Cell.
Mol. Biol.37:232-239.
Cao, D., Tal, T., Graves, L., Gilmour, I., Linak, W., Reed, W., Bromberg, P., and Samet, J. 2007.
Diesel exhaust particulate (DEP)-induced activation of stat3 requires activities of EGFR and
SRC in airway epithelial cells. American Journal of Physiology: Lung, Cell, & Molecular
Physiology, 292, L422-L429.
Cho, S.H., Yoo, J.-I., Turley, A.T., Miller, C.A., Linak, W.P., Wendt, J.O.L., Huggins, F.E., and
Gilmour, M.I. 2008.Relationships between composition and pulmonary toxicity of prototype
particles from coal combustion and pyrolysis, Proceedings of the Combustion Institute, 32, in
press.
Ciencewicki, J., Gowdy, K., Krantz, Q.T., Linak, W.P., Brighton, L., Gilmour, M.I., and Jaspers, I.
2007. Diesel exhaust enhanced susceptibility to influenza infection is associated with decreased
surfactant protein expression. Inhalation Toxicology, 19, 1121-1133.
77
-------
EPA CompTox Research Program FY2009 to 2012
Final
DeMarini, D.M., Brooks, L.R., Warren, S.H., Kobayashi, T., Gilmour, M.I., and Singh P. 2004.
Bioassay-directed fractionation an dsalmonella mutagenicity of automobile and forklift diesel
exhaust particles, Environmental Health Perspectives, 112, 814-819.
Gottipolu, R.R., Wallenborn, J.G., Karoly, E.D., Schladweiler, M.C., Ledbetter, A.D., Krantz,
Q.T., Linak, W.P., Nyska, A., Johnson, J.A., Thomas, R., Richards, J.E., Jaskot, R.H., and
Kodavanti, U.P. 2009. One-month diesel exhaust inhalation produces hypertensive gene
expression pattern in healthy rats. Environmental Health Perspectives. 117:38-46.
Gowdy, K., Krantz, Q.T., Daniels, M., Linak, W.P., Jaspers, I., and Gilmour, M.I. 2008.
Modulation of pulmonary inflammatory responses and anti-microbial defenses in mice exposed
to diesel exhaust. Toxicology & Applied Pharmacology, 229, 310-319.
Linak, W.P., Yoo, J.I., Wasson, S.J., Zhu, W., Wendt, J.O.L., Huggins, F.E., Chen, Y., Shah, N.,
Huffman, G.P., and Gilmour, M.I. 2007. Ultrafine ash aerosols from coal combustion:
characterization and health effects. Proceedings of the Combustion Institute, 31, 1929-1937.
Reed, W., Gilmour, I., DeMarini, D., Linak, W. and Samet, J. (2008). Gene expression profiles of
human airway epithelial cells exposed to diesel exhaust particles of varying composition. In
Preparation.
Saxena, RK, Williams, W & Gilmour, Ml. (2007) Suppression of basal and cytokine induced
expression of MHC, ICAM 1 and B7 markers on mouse lung epithelial cells exposed to diesel
exhaust particles. Am J Biochem Biotech. 3(4). 187-192.
Saxena, RK., Gilmour, Ml., & MD Hayes. 2007. Uptake of diesel exhaust particles by lung
epithelial cells and alveolar macrophages. Biotechnology 3 (4). 187-192
Singh, P., DeMarini, D.M., Dick, C.A.J., Tabor, D., Ryan, J., Linak, W.P., Kobayashi, T., and
Gilmour, M.I. 2004. Bioassay-directed fractionation, physiochemical characterization, and
pulmonary toxicity of automobile and forklift diesel exhaust particles in mice, Environmental
Health Perspectives, 112(8) 820-825.
Stevens, T., Krantz, Q.T., Linak, W.P., Hester, S., and Gilmour, M.I. 2008 Increased
Transcription of Immune and Metabolic Pathways in Naive and Allergic Mice Exposed to Diesel
Exhaust, Toxicological Sciences, 102(2), 359-370.
Stevens, T., and Linak, WP. Gilmour, Ml. 2009. Differential potentiation of allergic lung disease
in mice exposed to chemically distinct diesel samples. Tox Sci. 107(2), 522-534.
Stevens,T, Hester, S, & Gilmour Ml. Differential transcriptional changes in mice exposed to
chemically distinct diesel samples. Submitted.
Tal., T., Bromberg, P.A., Kim, Y. and Samet, J.M. (2008). Tyrosine phosphatase inhibition
induces epidermal growth factor receptor activation in human airway epithelial cells exposed to
diesel exhaust Toxicol. Appl. Pharmacol. 233:382-388.
Project Title: Development of Microbial Metagenomic Markers for Environmental
Monitoring and Risk Assessment
Peer-Reviewed Publications
78
-------
EPA CompTox Research Program FY2009 to 2012
Final
Lamendella, R., J. W. Santo Domingo, D. Oerther, J. Vogel, and, D. Stoeckel. 2007.
Assessment of fecal pollution sources in a small northern-plains watershed using PCR and
phylogenetic analyses of Bacteroidetes 16S rDNA. FEMS Microbiol. Ecol. 59:651-660.
Lamendella R, Santo Domingo JW, Yannarell AC, Ghosh S, Di Giovanni G, Mackie Rl, Oerther
DB. 2009. Evaluation of swine-specific PCR assays used for fecal source tracking and analysis
of molecular diversity of Bacteriodales-swine specific populations. Appl Environ Microbiol. 75:
5787-96.
Lamendella, R., Santo Domingo J.W., Kelty C, and Oerther DB. 2008. Occurrence of
bifidobacteria in feces and environmental waters. Appl. Environ. Microbiol. 74:575-584.
Lee, Y.J., M. Molina, and J.W. Santo Domingo, J.D. Willis, M. Cyterski, D.M. Endale, and O.C.
Shanks. 2008. A temporal assessment of cattle fecal pollution in two watersheds using 16S
rRNA gene-based and metagenome-based assays. Appl. Environ. Microbiol. 74:6839-6847.
Lu J, Santo Domingo JW, Hill S, Edge TA. Microbial Diversity and Host-specificSequences of
Canadian Goose Feces. Appl Environ Microbiol. 2009 Jul 24.
Lu, J. and J.W. Santo Domingo. 2008. Turkey fecal microbial community structure and
functional gene diversity revealed by 16S rRNA gene and metagenomic sequences. J.
Microbiol. 46:469-477.
Lu, J., J.W. Santo Domingo, R. Lamendella, T.Edge, and S.Hill. 2008. Phylogenetic diversity
and molecular detection of gull feces. Appl. Environ. Microbiol. 74: 3969-3976.
Lu, J., J.W. Santo Domingo, and O.C. Shanks. 2007. Identification of chicken-specific fecal
microbial sequences using a metagenomic approach. Water Res. 41:3561-3574.
Santo Domingo, J.W., D.G. Bambic, T.A. Edge, and S. Wuertz. 2007. Quo vadis source
tracking? Towards a strategic framework for environmental monitoring of fecal pollution. Water
Res. 41:3539-3552.
Santo Domingo, J.W. and T.A. Edge. 2009. Identification of primary sources of fecal pollution. In
Safe Management of Shellfish and Harvest Waters. G. Rees., K. Pond, D. Kay and J. Santo
Domingo. IWA Publishing, London, UK.
Shanks, O., J.W. Santo Domingo, J. Lu, C.A. Kelty, and J. Graham. 2007. PCR Assays for the
identification of human fecal pollution in water. Appl. Environ. Microbiol. 73: 2416-2422.
Shanks, O., J. W. Santo Domingo, R. Lamendella, C.A. Kelty, and J. Graham. 2006.
Competitive metagenomic DNA hybridization identifies host-specific genetic markers in cattle
fecal samples. Appl. Environ. Microbiol. 72:4054-4060.
Shanks, O., J. W. Santo Domingo, and J. Graham. 2006. Use of competitive DNA hybridization
to identify differences in the genomes of two closely related fecal indicator bacteria. J. Microbiol.
Methods. 66:321-330.
Vogel, J.R., D.M. Stoeckel, R. Lamendella, R.B. Zelt, J.W. Santo Domingo, S.R. Walker, and
D.B. Oerther. 2007. Identifying fecal sources in a selected catchment reach using multiple
source-tracking Tools. J. Environ. Qual. 36:718-729.
79
-------
EPA CompTox Research Program FY2009 to 2012
Final
Project Title: A Systems Approach to Characterizing and Predicting Thyroid Toxicity
Using an Amphibian Model
Peer-Reviewed Publications
Conners, K. , Jorte, J.J., Anderson G., Degitz SJ. Charaterization of thyroid hormone
transporting protein expression during tissue-specific metamorphosis in Xenopus tropicalis (In
review)
Degitz, S.J., Hornung, M.W., Korte, J.J, Holcombe, G.W, Kosian, P.A., Thoemke, K.R., Helbing,
C., Tietge, J.E. In vivo and in vitro regulation of genes in the thyroid gland following exposure to
the model T4 synthesis inhibitors methimazole, 6-propylthiouracil, and perchlorate. (In
preparation. To be submitted with above paper).
Hornung, M.W. Degitz, S.J., Korte, L.M., Olson, J., Kosian, P.A., Linnum, A.L., Tietge, J.E.
Inhibition of thyroid hormone release from cultured amphibian thyroid glands by methimazole, 6-
propylthiouracil, and perchlorate. (Completed, NHEERL In-House review. To be submitted with
the following Hornung et al. paper).
Hornung, M., Burgess, E. Tandem in vitro and ex vivo thyroid gland assays to screen xenobiotic
chemicals for thyroid hormone synthesis inhibition. (In preparation).
Sternberg, Thoemke, Hornung, Tietge, and Degitz. Regulation of thyroid-stimulating hormone
release from pituitary by t4 during metamorphosis in Xenopus laevis (In review)
Serrano, Higgins Witthuhn, Korte, Hornung, Tietge, and Degitz In vivo assessment and potential
diagnosis of xenobiotics that perturb the thyroid pathway: Part I. Differential protein profiling of
Xenopus Laevis brain tissues by two-dimensional polyacrylamide gel electrophoresis and
peptide-labeling with isobaric tags for relative and absolute quantification (iTRAQ) following
exposure to model T4 inhibitors. (In review).
Tietge Butterworth Kosian Hammermeister Hornung Haselman Degitz Analysis of Thyroid
Hormone and Related lodo-Compounds in Complex Samples by Inductively Coupled Plasma
Emission/Mass Spectrometry. (In preparation)
Tietge, Butterworth, Haselman, Holcombe, Korte, Kosian, Wolfe, Degitz. Early Temporal Effects
of Three Thyroid Hormone Synthesis Inhibitors in Xenopus laevis. (In preparation)
Project Title: Mechanistic Indicators of Childhood Asthma (Mica): A Systems Biology
Approach For the Integration of Multifactorial Environmental Health Data
Peer-Reviewed Publications
Kim SJ, Dix DJ, Thompson KE, Murrell RN, Schmid JE, Gallagher JE, Rockett JC. 2007. Effects
of storage, RNA extraction, genechip type, and donor sex on gene expression profiling of
human whole blood. Clin Chem. Jun;53(6):1038-45.
Vesper, S.,McKinstry C., Haugland., R., Neas, L., Hudgens, E., Heidenfelder, B., and Gallagher
J. Environmental Relative Moldiness Index (ERMIsm) as a Tool to Identify Mold Related Risk
Factors for Childhood Asthma Sci Total Environ. May 1;394(1): 192-6 (2008)
Johnson M, Hudgens E, Williams R, Andrews G, Neas L, Gallagher J, Ozkaynak H. 2009. "A
Participant-Based Approach to Indoor/Outdoor Air Monitoring in Community Health Studies"
Journal of Exposure Science and Environmental Epidemiology. 19(5):492-501.
80
-------
EPA CompTox Research Program FY2009 to 2012
Final
Cohen Hubal E, Richards A., Shah I, Edwards S, Gallagher J, Kavlock R, Blancato, J. 2008.
Exposure Science and the US EPA National Center for Computational Toxicology J Expo Sci
Environ Epidemiol.. November [Epub ahead of print],
Heidenfelder B,. Reif D, Harkema, JR, Cohen Hubal E, Hudgens,E. Bramble L G. Wagner G,
Harkema JR, Morishita M, KeelerG , Edwards,SW and Gallagher J. 2009. Comparative
Microrarray Analysis and Pulmonary Changes in Brown Norway Rats Exposed to Ovalbumin
and concentrated Air Particulates Tox Sci. 108(1 ):207-21.
Heidenfelder B, Johnson M, Hudgens E, Inmon J, Hamilton R, Neas L, and Gallagher J,
Increased plasma reactive oxidant levels and their relationship to blood cells, total IgE, and
allergen-specific IgE in asthmatic children. 2009. Journal of Asthma (accepted).
Williams AH, Gallagher JE, Hudgens E, Johnson MM, Mukerjee S, Ozkaynak H, Neas LMN.
EPA Observational studies of children's respiratory health in Detroit and Dearborn, Michigan.
Proceedings of AWMA 102nJune 16-19; Detroit, Michigan.2009.
J. E Gallagher, E A Cohen Hubal, S.W. Edwards Invited book Chapter "Biomarkers of
Environmental Exposure" "Biomarkers of toxicity: A New Era in Medicine Editors Vishal S.
Vaidya and Joseph V. Bonventre Publisher:John Wiley and Sons, Inc. October 1, 2009.
Markey M. Johnson, Ron Williams, Zhihua Fan, Lin, Edward Hudgens, Jane Gallagher, Alan
Vette, Lucas Neas, Haluk Ozkaynak Indoor and outdoor concentrations of nitrogen dioxide,
volatile organic compounds, and polycyclic aromatic hydrocarbons among MICA-Air households
in Detroit, Michigan submitted AWMA 2009.
Gallagher, J Reif, D; Heidenfelder, B Neas, L; Hudgens, E Williams, A Inmon, J; Rhoney, S,
Andrews G., Johnson, M Ozkaynak, H; Edwards, S, Cohen-Hubal, E. 2009. Mechanistic
Indicators of Childhood asthma ( MICA); A systems biology approach for the integration of
multifactorial environmental health data. Journal of Exposure Science and Environmental
Epidemiology, (submitted).
81
------- |