United States
Environmental Protection
Agency
Perspectives on the Development, Evaluation, and
Application of in Silico Approaches for Predicting
Toxicity
* 4 1
¦/••••• ;n.
v. *<•
Grace Pa+lewicz
Center for Computational Toxicology and Exposure (CCTE), US EPA
The views expressed in this presentation are those of the author and do not necessarily reflect the views or policies of the U.S. EPA
-------
•"¦pRfk Conflict of Interest Statement
Environmental Protection
Agency
No conflict of interest declared.
Disclaimer:
The views expressed in this presentation are those of the authors and do not
necessarily reflect the views or policies of the U.S. EPA
National Center for
Computational Toxicology
1
-------
United States
Environmental Protection
Agency
Outline
• Regulatory Drivers
• Computational (in silico)Toxicology [scope for today's talk]
•Integrated Approaches to Testing and Assessment (IATA) - definitions
and Adverse Outcome Pathway (AOP) informed
• Decision contexts and their impact on the approaches applied
• Risk-based prioritization
-Thresholds for Toxicological Concern (TTC)
• Read-across approaches
-Generalised Read-across (6enRA)
- Perfluorinated &polyfluorinated substances (PFAS)
• Summary remarks
• Acknowledgements
National Center for
Computational Toxicology
-------
Regulatory and Non-Regulatory drivers
United States w
Environmental Protection
Agency
• Societal demands for safer and sustainable chemical products are
stimulating changes in toxicity testing and assessment frameworks
• Chemical safety assessments are expected to be conducted faster and
with fewer animals, yet the number of chemicals that require
assessment is also rising with the number of different regulatory
programmes worldwide.
• In the EU, the use of alternatives to animal testing is promoted.
• Animal testing is prohibited in some sectors e.g. EU Cosmetics
regulation
• The European Registration, Evaluation, Authorisation and Restriction
of Chemicals (REACH) legislation lays out specific information
requirements, based on tonnage level triggers. However, the regulation
explicitly expresses the need to use non-testing approaches to reduce
the extent of experimental testing in animals.
National Center for ¦
Computational Toxicology
-------
Regulatory and Non-Regulatory drivers
United States w w
Environmental Protection
• BREACH-like schemes also have been established in China, South Korea,
and Turkey.
• In the US, the new Frank Lautenberg Chemical Safety for the 21st
Century Act (LCSA) requires that a risk based prioritisation is
conducted for all substances in commerce, ~40,000, many of which
are lacking sufficient publicly available toxicity information.
• EPA Administrator signed memo 10/9/19 to "direct the agency to
aggressively reduce animal testing, including reducing mammal study
requests and funding 30% by 2025 and completely eliminating them by
2035"
• Risk based prioritisation is also an important aspect of regulatory
frameworks in Canada (the Domestics Substance List), Australia and
the EU.
• Non-testing approaches offer a means of facilitating the regulatory
jcl^gJl^nges in chemical safety assessment
Computational Toxicology
-------
United States
Environmental Protection
Agency
Computational (In Toxicology
Non-Testing Approaches
Databases/Dashboards of existjj
Structure-
Quantitativ
Expert Sys
Category f<
Bioinformatic
Chemoinformatics
Biokinetics (PBPK)
R)
National Center for
Computational Toxicology
-------
«>EF¥\ Integrated Approaches to Testing and
Environmental Protection
Assessment (IATA)
•"Integrated Testing Strategies (ITS) are .... approaches that
integrate different types of data and information into the
decision-making process. ..."
• "A tiered approach to data gathering, testing, and assessment that
integrates different types of data (including physicochemical and
other chemical properties as well as and toxicity
data). When combined with estimates of exposure in an appropriate
manner, the IATA provides predictions of risk."
National Center for
Computational Toxicology
-------
<>EPA General framework of an I ATA
United States
Environmental Protection
Agency
Available information
provides sound
conclusive evidence for
the specific regulatory
need
Make a weight of evidence assessment or apply predefined decision
criteria (e.g. ITS, STS).
Gather and evaluate existing information (in vivo, in vitro, in silico
(e.g. (Q)SAR), read across and chemical category data).
Problem formulation. Definition of the regulatory need (e.g. hazard
identification, hazard characterisation, safety assessment etc.) and
the information/parameters that are relevant to satisfy the need,
including consideration of existing constraints and, if applicable,
consideration of the level of certainty required.
If available information does not provide sufficient evidence
consider what additional information from non-testing, non-animal
testing methods and, as a last resort, from animal methods would be
needed to generate sufficient evidence.
Make a weight of evidence assessment or apply predefined decision
criteria (i.e. ITS, STS).
National Center for
Computational Toxicology
Available information
provides sound
conclusive evidence for
the specific regulatory
need
From OECD 7
-------
United States
Environments
Agency
Environmental Protection
Typical Information within an IATA:
I ATA elements
• Historical information on the chemical of interest
• Non-standard in vivo tests
• Information from "similar" chemicals
• Predictions from other 'non-testing' approaches such as (Q)SAR
• In chemicotests
• In vitro tests
• Molecular biology, -omics
• Exposure, (bio-)kinetics
National Center for
Computational Toxicology
-------
4>EPA Mechanistic based and AOP-informed IATA
United States
Environmental Protection
Agency
• As noted, there is o shift towards non animal alternatives as a response to
regulatory drivers
• Integration of different non-animal approaches requires an organising framework to
ensure that the different information sources are being interpreted in their
appropriate context. This is particularly relevant for New Approach Methodologies
(NAMs).
• AOPs serve to provide this organisational framework and hence play an important
role in developing and applying IATA for different purposes as well as provide a
roadmap for future QSAR development
• AOPs provide the linkage from chemistry, through the Molecular Initiating Event
(AAIE) to Adverse Effect
• Data from key events provides support to, and will enhance, read-across especially
for regulatory acceptance as well as supports definition of domains for MIEs
National Center for
Computational Toxicology
-------
<>EPA General workflow in Integrated Approaches to Testing and
United States I I
Environmental Projection
Agency
Assessment (IATA)
Problem formulation
informa
on-making?
Generate additional information
YES
Regulatory
conclusion
YES
National Center for
Computational Toxicology
Weight of Evidence
assessment: Adequate
information for decision-making?
From OECD
-------
EPA CompTox Chemicals Dashboard
Environmental Protection
Agency
A publicly accessible website delivering access:
-~875,000 chemicals with related property data
-Experimental and predicted physicochemical property data
-Integration to "biological assay data" for 1000s of chemicals
-Information regarding consumer products containing chemicals
-Links to other agency websites and public data resources
-"Literature" searches for chemicals using public resources
-"Batch searching" for thousands of chemicals
- DOWNLOADABLE Open Data for reuse and repurposing
National Center for
Computational Toxicology
https://comptox.epa.gov/
-------
United States
Environmental Protection
Agency
CompTox Chemicals Dashboard:
:ion
Landing Page
vvEPA
United States
Environmental Protection
Agency
Advanced Search Batch Search
Chemicals
875 Thousand Chemicals
Product/Use Categories Assay/Gene
1J Identifier substring search
See what people are saying, read the dashboard comments!
Cite the Dashboard Publication click here
Latest News
Read more news
New Article regarding the GenRA module
March 9th, 2019 at 1:03:58 PM
A new article regarding "Generalized Read-Across (GenRA): A workflow implemented into the EPA CompTox Chemicals Dashboard" has been
published in theALTEX (Alternatives to Animal Experimentation) journal. Read the article here.
National Center for
Computational Toxicology
-------
-------
vvEPA CompTox Chemicals Dashboard:
rejection | « A J* • m • § * I
Landing Page for a specific chemical
United States
Environmental Projection
Agency
v>EPA
United States
Environmental Protection Home Advanced Search Batch Search Lists v Predictions Downloads
Agency
EXECUTIVE SUMMARY
PROPERTIES
ENV. FATE/TRANSPORT
HAZARD
~ ADME
~ EXPOSURE
~ BIOACTIVITY
SIMILAR COMPOUNDS
GENRA (BETA)
RELATED SUBSTANCES
SYNONYMS
~ LITERATURE
LINKS
COMMENTS
Bisphenol A
80-05-7 I DTXSID7020182
Searched by DSSTox Substance Id.
V
Wikipedia
Bisphenol A (BPA) is an organic synthetic compound with the chemical formula (CHs^Ca^OHk belonging to the group of diphenylmethane
derivatives and bisphenols. with two hydroxyphenyl groups. It is a colorless solid that is soluble in organic solvents, but poorly soluble in water. It has
been in commercial use since 1957.
BPA is a starting material for the synthesis of plastics, primarily
Intrinsic Properties
Structural Identifiers
Linked Substances
Presence in Lists
Record Information
Quality Control Notes
National Center for
Computational Toxicology
-------
CompTox Chemicals Dashboard:
Environmental Protection I
Agency
Executive Summary
v>EPA
United States
Environmental Protection Home Advanced Search Batch Search Lists v Predictions Downloads
Agency
EXECUTIVE SUMMARY
PROPERTIES
ENV. FATE/TRANSPORT
HAZARD
~ ADME
~ EXPOSURE
BIOACTIVITY
TOXCAST: SUMMARY
EDSP21
TOXCAST/TOX21
PUBCHEM
Bisphenol A
80-05-7 | DTXSID7020182
Searched by Expert Validated Synonym.
Executive Summary
Quantitative Risk Assessment Values
© IRIS values available C?
Q No PPRTV values
© EPA RSL values available G?
O Minimum RfD: 0.050 mg/kg-day (chronic, IRIS, oral, 8) Gf
Q Mo RfC calculated
Q IVIVE POD not calculated
Quantitative Hazard Values
Q Minimum oral POD: 3.8 mg/kg-day (reproductive, HPVIS, oral, 6) C?
© No inhalation POD values
Q Lowest Observed Bioactivity Equivalent Level: CYP1A1, CYP1A2, Tpo, ESR2, ESR1,
ESR1, NR1I3, PPARA, NR1I2, Cyp2c11, MMP3, Esr1
Cancer Information
© No cancer slope factor
© No inhalation unit risk value
(j) Carcinogenicity data available: University of Maryland carcinogenicity warning; G?
No genotoxicity findings reported
Reproductive Toxicology
Q 200 Reproductive toxicity PODs available Gf
Class
risk-based SSL (mg/kg)
GIABS (unspecified)
GIABS (unspecified)
ABS (unspecified)
RFDo (mg/kg-day)
screening level (residential Soil) (mg/kg)
screening level (industrial soil) (mg/kg)
REGIONAL SCREENING
THQ
THQ = 0.1
THQ = 1
THQ = 0.1
THQ = 0.1
THQ = 0.1
THQ = 0.1
THQ = 0.1
Value
5.8
1
1
0.1
0.05
320
4100
National Center for
Computational Toxicology
-------
v>EPA Computational toxicology tools add value
United States I 9
Environmental Protection m ¦ • •
to most regulatory decisions
• Prioritisation
•Screening level hazard assessment
• Risk Assessment
• Exposure Assessment
National Center for
Computational Toxicology
-------
«>EF¥\ Risk-Based prioritisation
Environmental Protection
Agency
* Could involve a combination of available experimental data and new
approach methods (NAMs) such as HTTR, HTS
* One approach considered involved coupling Threshold of Toxicological
Concern (TTC) with High Throughput Exposure (HTE) modelling to
rank order substances for further evaluation
* TTC is a principle that refers to the establishment of a human
exposure threshold value for (groups of) chemicals below which there
would be no appreciable risk to human health
* Relies on past accumulated knowledge regarding the distribution of
potencies of relevant classes of chemicals for which good toxicity
data do exist
TTC is based on a predicted tumour risk of 1 in a million, derived through on
analysis of genotoxic chemicals
TTC is based on frequency distributions (5th percentile) of NO(A)ELs of non-
genotoxic chemicals
National Center for
Computational Toxicology
-------
<>EPA TT values
United States
Environmental Protection
Agency
Type of substance
Alerts for potential genotoxic
carcinogenicity
Acetylcholinesterase inhibitors
(AChEl)
Organophosphate/carbamate
Cramer Class III
Cramer Class II
Cramer Class I
National Center for
Computational Toxicology
gg/person/day (jjg/kg-day for 60 kg
adult)
Kroes: 0.15 (0.0025 ug/kg-day)
ICH: 1.5 (0.025 ug/kg-day)
18 (0.3 ug/kg-day )
90 (1.5 ug/kg-day)
540 (9.0 ug/kg-day)
1800 (30 ug/kg-day)
-------
United States
"I Protection
Cumulative Distributions of Cramer Structural
Class NOELs
Decision tree of 33 questions
[20H I0).(23) m
27 22 21
25 16 Kt " 28 33 H 18 m 2 2
/\
26 n I 31 30. 29 m I I H 33 II HI '4
13
22 H
33 XL
hi i
m i
22
15
33
m I 187(19) 31 33 H - - 33 22,(16.)
i n 32 18,(19) mi mr i 33 n
y-v
22 n 1 n in 1
in 1
O
s_
fc)
Q_
100
90
80
70
60
50
40
30
20
10
NOEL (mg/kg/day)
Fitted
Distribution
Class I
Class II 7
Class IIP
fi-
1
1 1
1 1 1
1 1 1 III
1 1
1 1 1 1 1 11
—¦—
1 1 1 1 1 111
1 1 1 1 1 1 111
-e\
01
0.1
1.0
10
100
1000
10000
National Center for
Computational Toxicology
-------
Predicted HT
Environmental Protection
Agency
* Wambaugh and colleagues (2014) developed a rapid heuristic
high throughput exposure (HTE) model that enables prediction
of potential human exposure to thousands of substances for
which little or no empirical exposure data are available.
• The HTE model was calibrated by comparison to NHANES
urinary data that reflects total exposure (all routes/sources)
exposures
National Center for
Computational Toxicology
-------
United States
Environmental Protection
Agency
Integrating TT Cwith predicted HT
ction I
exposures
Compared the conservative Cramer Class III TTC value of 1.5 pg/kg-
day to the previously calculated median and upper 95% credible interval
(UCI) of total daily median exposure rates for 7968 chemicals
TTC = 0.0015 mg/kg/d
0.4 0.6
Cumulative Frequency
only 273 (fewer than 5%) were found
to have UCI daily exposures estimates
that exceeded the Cramer Class III
TTC value of 1.5 jjg/kg-day
Initial evaluation showed the approach of using
the ratio of exposure to TTC (HTE: TTC)
appeared promising for risk-based prioritisation
National Center for
Computational Toxicology
-------
•»ER^ Risk-Based prioritisation
Environmental Protection
Agency
• Refined the approach using the Kroes et al structure-based workflow
for TTC
Is the substance a non-essential metal Or metal containing compound, c
-dibcnzofuran, or -biphenyl
a polyhalogenated-dibeiuodioxin,
| 2. Are there structural alerts that raise
ti foe potential Renotoxicity?
5. Does estimated intake exceed 1
TTC of l-5psfclay?
YES
NO |
Substance would not be
; expected to be a safety concern •
! Risk assessment requires compound-specific
toxicity data
| 6. Is the compound an organophosphatei
Does estimated intake exceed TTC
of O.ISpg/day?
Negligible risk (low probability of a life-t
cancer risk greater than 1 in lO*" -
7. Does estimated intake exceed
TTC of 18ug/day
9. Does estimated intake exceed
90pg/day
requires
compound-specific toxicity data
10. Is the compound
al class II?
: Substance Mould not be expected
to be a safety concern
11. Does estimated intake exceed
540ii g/day
12. Does estimated intake
exceed 180Oug/day?
Risk assessment requires
compound-specific toxicity data
Substance would not be expected
be a safety concern
Determines if TTC
approach is
appropriate based on
chemical structure
Is the substance one
these types?
Stffc«PCDOF*
• metal or
MfariMnetafcc
a nanwnatantf
i raicKhe
substance
win
ol
* 9K0>d
Perform
Substance-specific
Risk Assessment
Structural Alerts for
Gen otoxicity identified
by QSAR-based
rules?
Alerts for high potency
genotoxic compounds
(Aflatoxin-lifce, azoxy-
or N-nitroso)?
Compare to the range
of TTC levels based
on chemical structure
Candidates for
further evaluation of
hazard and exposure
Genotoxic
Cbemcais f 015ii-5>y»>
(Assumed ft M2S pj-VVdsi^
Carcinogens}
AriD-Ctts (QP3 18
ar*d Carbamate# 03 ugAgMnf
9U uiOif
urararaass
1 & paf«a'saj'
540 pyday
C"3T«r Class
I8DC yV.iiy
Cramer Class I
3D LiiVvdiv
Compare and
Rank HT
Exposures to
TTC Values
Lower priority for
further evaluation at
this time
None of the substances categorised as Cramer Class I or Cramer Class II exceeded their respective TTC
values.
No more than 2% of substances categorised as Cramer Class III or acetylcholinesterase inhibitors exceeded
their respective TTC values.
Majority of chemicals with genotoxicity structural alerts did exceed the relevant TTC - recommendations were
proposed for next steps
National Center for
Patlewicz et al, 2018
Computational Toxicology
-------
Risk-Based prioritisation
Environmental Protection
Agency
Investigate relevance of existing TTC values for substances of
interest to EPA
Extracted data from EPA's ToxVal DB, which aggregates
testing data from over 40 sources including US federal and
state agencies, as well as international agencies such as the
European Chemicals Agency and the World Health Organisation
Objectives were:
• Reproduce the TTC values developed by Munro et ol (1996)
• Follow the Kroes et ol (2004) workflow to assign substances present in
ToxVal to their respective Cramer classes and use the associated
repeat dose toxicity data to derive new TTC values
• Evaluate whether the TTC values from ToxVal and AAunro are
statistically equivalent
• Derive confidence intervals for the new TTC values
• Compare and contrast the chemistry of the two data sets to
rationalise any (dis)similarities in TTC values
National Center for
Computational Toxicology
-------
v>epa Risk-Based ppioritisation
I I m ¦ + '—¦» O -V #-» 4- j-*
United States
Environmental Protection
Agency
Follow the Kroes et al (2004) workflow to assign
substances present in ToxVal to their respective Cramer
classes and use the associated repeat dose toxicity data
to derive new TTC values
Evaluate whether the TTC values from ToxVal and
Munro are statistically equivalent derive confidence
intervals for the new TTC values
CDFs of Cramer Classes
from ToxVal data
. I , I.M.I I .
0 1 2
NO(A)EL Log10(mg/kg bw/day)
Class I
-•-Class II
-•-Class III
j .
Cramer Class I
Cramer Class II
Cramer Class III
90%
80%
70%
20%
10%
90%
80%
70%
60%
50%
40%
30%
20%
10%
i 'j *. V :t
NO(A)EL Log10(mg/kg bw/day)
I . .
o 1
NO(A)EL Log10(mg/kg bw/day)
National Center for
Computational Toxicology
ToxVal
Munro
I I I I I
0 12 3 4
NO(A)EL Log10(mg/kg bw/day)
-------
*»1PA Risk-Based prioritisation
Environmental Protection I
Agency
• Bootstrap sampling used to quantify the uncertainty around the 5th
percentiles values for both ToxVal and AAunro data sets
• Differences were observed for substances assigned as Cramer Class
III
<
O
Class II
Cramer Class
• Presence of OP/carbamates in the AAunro Cramer class III set largely
explained the difference in 5th percentile values
National Center for
Computational Toxicology
Nelms et al, submitted
-------
v>EPA
United States
Environmental Protection
Agency
Definitions: Chemical grouping
approaches
"Analogue approach" refers to grouping based on a very limited number of chemicals
(e.g. target substance + source substance)
"Category approach" is used when grouping is based on a more extensive range of
analogues (e.g. 3 or more members)
A chemical category is a group of chemicals whose physico-chemical and human
health and/or environmental toxicological and/or environmental fate properties are
likely to be similar or follow a regular pattern as a result of structural similarity (or
other similarity characteristics).
National Center for
Computational Toxicology
-------
United States
Environmental Protection
Agency
Definitions: Read-across
Read-across describes the method of filling a data gap whereby a
chemical with existing data values is used to make a prediction for a
'similar' chemical.
A target chemical is a chemical which has a data gap that needs to
be filled i.e. the subject of the read-across.
A source analogue is a chemical that has been identified as an
appropriate chemical for use in a read-across based on similarity to
the target chemical and existence of relevant data.
Source
chemical
Target
chemical
Property
cs=
•
• Reliable data
° Missing data
Acute
toxicity?
Known to be
harmful
Predicted to be
harmful
National Center for
Computational Toxicology
-------
v>epa A harmonised hybrid read-across workflow
United States m
Environmental Protection
Agency
ELSEVIER
Contents lists available at ScienceDirect
Computational Toxicology
journal homepage: www.elsevier.com
Journal
Image
Navigating through the minefield of read-across frameworks: A commentary
perspective
Grace Patlewicza' *, Mark T.D. Croninb, George Helmana'c, Jason C. Lambert^ Lucina E. Lizarragad, Imran Shah2
a National Center for Computational Toxicology (NCCT), Office, of Research and Development, US Environmental Protection Agency (US EPA), 109 TW Alexander Dr, Research Triangle Park
(RTP), NC 27711, USA
b School of Pharmacy and Biomolecular Sciences, Liverpool John Moores University, Byrotn Street, Liverpool L3 3AF, UK
c Oak Ridge Institute for Science and Education (ORISE), 1299 Bethel Valley Road, Oak Ridge, TN 37830, USA
d National Center for Evaluation Assessment (NCEA), US Environmental Protection Agency (US EPA), 26 West Martin Luther King Dr, Cincinnati, OH 45268, USA
National Center for
Computational Toxicology
29
-------
1 Protection •
betermine the scope of the
assessment needed
e.g. screening level hazard
assessment
Determine number and
type of data gaps
i. Decision
context
The number of data gaps and fop which
endpoints will drive the approach to fill
the data gaps, «e,g. using defined
approaches or QSARs
Custom search specific to endpoint specific
parameters OR
Search on the basis of structural similarity and/or
other similarity contexts to address a broader
number of endpoints
Evaluate on the basis of physchem, metabolism,
reactivity, TK, toxieologieal.ete
Also evaluate consistency and concordance of
experimental data (both effects and potency) of the
source analogues -across the endpoint, between
endpoints (temporal and dose response relationship)
and relative to the target using the data matrix .
Assess prediction and uncertainty relative
(prediction uncertainty and underlying data
variability) to the decision context (5hah et al
(2016)- refine analogue identification as required
Generate new information depending on the sources
of the uncertainties see Patlewicz et al (2015) it
Schultz et al (2015)
Consider Defined
Approaches in the
context of an I ATA
2. Data gap
analysis for
target
endpoint
pathway
Is/are the data
gap(s) for
physicochemical
ecotox or e-fate
properties?
Consider QSAR
approaches
sensitization
oestragemcity
0 ^
3. Overarching
similarity rationale
n
Rationale(s) are either more broadly
defined on the basis of functional
groups, reactivity etc. or specific to
an endpoint
4. Analogue
identification
o
5. Analogue
evaluation
JT
Q
6. Data gap filling
1
TXJ
Qualitative/Quantitative read-across.
Trend analysis. External QSAR
7. Uncertainty
assessment
Where do other NAM fit?
How should we transition to
data-driven approaches?
What about characterising the
uncertainty of the predictions
made?
Fig. 9. A harmonised hybrid development and assessment framework.
National Center for
Computational Toxicology
Patlewicz et al., 2018
-------
United States
Environmental Protect!
Agency
Selected read-across tools
Computational Toxicology 3 (2017) 1-18
ELSEVIER
Content* lists available at SciencsDirect
Computational Toxicology
journal homepage: www.elsevier.eorWIocaie/comtox
Navigating through the minefield of read-across tools: A review of in
silico tools for grouping
Grace PatlewiczGeorge Helman3-13, Prachi Pradeepa-b, Imran Shah3
¦Nurioniri Center for CoinputcfJOJEaJ Toxicology (NCCTJ, Office of Research and Development, US Lnvironmrncat iTotertion Agency,
)09 TW Alexander Dr, Jteseardl Inangte mfc (RTF), NC27711, USA
b0at Ridge Imtiiutefar Science and tduiuooK fGEKil, Oafe Ridge, m, L'M
CrossMs rk ^
ARTICLE
N F O
Keywords:
Category approach
Analogue approach
Data gap filling
Read-across
-------
Selected read-across tools
United States
Environmenta
Agency
iminn
Tool
AIM
ToxMatch
AMBIT
OECD
Toolbox
CBRA
ToxRead
GenRA
Analogue
identification
X
X
X
X
X
X
X
Analogue N A X X X X X
Evaluation by For
other Ames &
tools BCF
availabl
e
NA
Data gap
analysis
N A
X
X
Data
matrix
can be
exporte
d
X
Data
matrix
viewable
N A
N A
X
Data
matrix can
be
exported
Data gap NA X User XXX
filling driven
X
Uncertainty
assessment
N A
N A
N A
X
N A
NA
X
Availability Free Free Free Free Free Free
Free
-------
SenRA (Generalised Read-Across)
Agency
•Predicting toxicity as a similarity-weighted activity of
nearest neighbours based on chemistry and bioactivity
descriptors (Shah et al, 2016)
•Generalised version of the Chemical-Biological Read-Across
(CBRA) developed by Low et al (2013)
•Goal: To establish an objective performance baseline for
read-across and quantify the uncertainty in the predictions
made
a U{chm, bio , be)
lkSaXP Jaccard similarity: pn{6fo,/Qr}
"V P ~~ ^^_ ^/( Xil A jV ) v,= predicted activity of chemical(Cj)
C( ="7 7
L : liiXuVXj,)
J V
National Center for
Computational Toxicology
Xy= activity of c fm. (B
s° = Jacccard similarity between xfa, xj
k- up to k nearest neighbours
-------
«iPA GenRA vl.O - Approach
">1 Protection
1,778 Chemicals
3,239 Structure descriptors
(chm)
820 Bioactivity hitcalI (bio)
ToxCast
574 toxicity effects (tox)
ToxRefDB
CHR
SUB
MGR
Liver
Kidney
Spleen
Adrenal Gland
Lung
Thyroid Gland
Testes
Stomach
Brain
Heart
Ovary
Eye
Uterus
Bone Marrow
Lymph Node
Pituitary Gland
Thymus
Skin
Pancreas
Mammary Gland
Urinary Bladder
Epididymis
intestine smah
Blooa
Bone
Intestine Large
Parathyroid Gland
Skeletal Muscle
Nerve
Gallbladder
Seminal Vesicle
Salivary glands
Harderian Gland
Spinal cord
Trachea
Ear
Blooc vessel
Parathyroid
Vagina
Esophagus
Oral Mucosa
Penis
Lacrimal Gland
Mesentery
Coorci nation
Larynx
Placenta
Reflexes
Ureter
§§§111
: T"?
II. Define Local
neighbourhoods
Use K-means analysis to
group chemicals by similarity
Use cluster stability analysis
~ 100 local neighbourhoods
DEV
§§§1 = 1
n_pos
n_neg
c:i Htj rhm C
Computational Toxicology
-------
United States
Environmental Protection
Agency
Read-across workflow in GenR vl.O
Decision Context
Screening level assessment of
hazard based on toxicity effects
from ToxRefDB vl
Uncertainty
assessment
Assess prediction and
uncertainty using AUC and p
value metrics
Analogue
identification
Similarity context is based on
structural characteristics
Read-across
Similarity weighted average
many to one read-across
*
Data gap analysis
for target and
source analogues
Analogue evaluation
Evaluate consistency and
concordance of experimental
data of source analogues across
and between endpoints
National Center for
Computational Toxicology
-------
United States
Environmental Protection
Agency
GenRA tool in reality
Integrated into the EPA CompTox Chemicals dashboard
Neighbors by: Chem: Torsion Fgrprts T Filter by: invivo data T
Summary Data Gap Analysis
Group: ToxRef
By: Tox Fingerprint
Generate Data Matrix
A
3-Phenylprop-2-enal
7
187
14
168
CHR:Bile duct
2-Chloroacetophenone
20
Q
13
168
CHRBipod
(+/-)-alpha-Methylbenzyl...
21
271
8
168
CHR:Blood vessel
Ctorophene
m
13
3,
CHR:Body Wejgtit
Monobenzyl phthalate
13
9
95
CHR: Bone
Benzophenone
2
m
9
Q
CHR:Bone Marrow
Naphthalene
8
m
2
178
CHR:Brain
J' & £,¦ <0> P; lT
<0 -3-F?ienytp.- 2-CWoroacetop.. t>:-}-alpha-Met. Ctorophene Monobenzyl ph. Benzophenone Naphthalene
CHR;Abdommal Cavity
CHR Adrertal Gland
CHR:AftefY.(Gttieralj
CHR Auditory Startle Re. .
CHR:Bileduct
CHR: Blood
CHR: Blood vessel
CHRBody Weight
CHR: Bone
National Center for
Computational Toxicology
-------
United States
GenRA tool
• Structured as a workflow
^1 Protection
in practice
DETAILS
EXECUTIVE SUMMARY
PROPERTIES
ENV. FATE/TRANSPORT
HAZARD
ADME
> EXPOSURE
> BIOACTIVITY
SIMILAR COMPOUNDS
RELATED SUBSTANCES
SYNONYMS
> LITERATURE
LINKS
COMMENTS
Fluconazole
86386-73-4 | DTXSID3020627
Searched by DSSTox Substance Id.
Step One: Analog Identification and Evaluation
Similarity context
CH
Bromucorazole
tjoiriputafionrfftoxicology
Cypfoconazole
Pyrasutfotc-:© m...
V
y-o-
xk-
Mycobutanil
# of Analogs 10
Tetraconazc-te
Fenbuoorrazole
-A.
i
4
Neighbors by: Achem: Morgan Fgrprts ~
niter by:
invivo data ~
0
V
y
¦*
-i
i
4
i
4
-------
GenRA tool in practice
GenRA
Step Two: Data Gap Analysis & Generate Data Matrix
Neighbors by: Cherri: Morgan Fgrprts *
Ethylene glycol...
Filter by: invivo data *
Sum
il^ry
Data Gap Analysis
a
Group: ToxRef
<>
<7/
*r i
Butanal oxime
Acrolein diethyl..
Elhoprop
r *
jz.
Fosamine amm..
# of Analogs 10
„^C/=C
Myrcene
*} .
.Wc
c
Chlorethoxyfos
Methyl eugenol
bis(2-Chloro-1-
2-Elhoxyethyl a...
m
^Fluconazole
3
714
15
0
Hexaconazole
43
819
18
345
Flusilazole
28
819
9
345
Cyproconazole
14
819
16
408
Pyrasulfotole metabolite ...
0
0
18
234
Myclobutanil
15
818
15
345
Fenbuconazole
34
819
17
345
Tetraconazole
35
819
20
345
Metconazole
35
215
15
82
Ipconazole
46
232
16
180
Bromuconazole
24
277
13
345
Data gap analysis
By: Tox Fingerprint »
Generate Data Matrix
&:
/ Z Z / #' /
*/ / # €' / #
/ #¦ /!'¦ /' //
c« o i fT' 5* a •
.6' IX, Sv -0; 8
/ ^ ,f J
# ^ cf
8 P
-------
United States
GenRA tool in practice
GenRA
Neighbors by: Chem: Morgan Fgrprts ~ Filter by: invivo data ~
Ethylene glycol..
Step Three: Run GenRA Prediction
Summary Data Gap Analysis
ToxRef ~
By:
Tox Fingerprint *
Run Read-Across
,
# J*-
<£¦
M / / i/ S $/ / i/ /
si
®, ¦ c ® -f e>/ s> „ o*
//1 */ i i/ Jr w £; i £¦ /
^ «cr A. ^ &/ S J?
m 4/ # $ # if «° <£
Butanal oxime
Myrcene
-&¦ S.:-' -
i*Y \ !-.• t_»
Hexaconazole Flu&lazole Cyproconazoie Pyrasu Ifotole m... Myclofoutani Fenbuconazoie Tetraconazofe Metconazole Ipconazo-Se Bromuconazole
CHR: Abdominal Cavity
CHR:Adrenal Gland
CHR:Artery (General)
CHR.Auditory Sfarie Re..-
CHR.:Bile duct
CHRBIood
CHR:BSood vessel
CHR: Body Weight
CHR:Bone
National Center for
Computational Toxicology
-------
United States
Neighbors by: Chem: Morgan F
Butanal oxime
Run GenRA
National Center for
Computational Toxi<
GenRA tool in practice
ALTEX preprint
published February 4, 2019
doi:10.14573/aitex. 1811292
Short Communication
Generalized Read-Across (GenRA): A workflow
implemented into the EPA CompTox Chemicals
Dashboard
George Helmarv % Imran Sha'ir, Antony J. Williams2, Jeff Edwards2, Jeremy Dunne2 and Grace
Patlewicr
!Oak Ridge Institute for Science and Education (ORISE). Oak Ridge, TN, USA; -National Center for Computational
Toxicology (NCCT), Office of Research and Development US Environmental Protection Agency, Research Triangle Park
(RTP). KC: USA
Abstract
Generalized Read-Across (GenRA) is a data driven approach which makes read-across predictions on the basis of a
similarity weighted activity of source analogues (nearest neighbors). GenRA has been described in more detail in the
literature (Shah et al.. 2016; Helman et aL 2013). Here we present its implementation within the EPA's CompTox
Chemicals Dashboard to provide public access to a GenRA module structured as a read-across workflow. GenRA
assists researchers in identifying source analogues, evaluating their validity and making predictions of in vivo toxicity
effects for a target substance. Predictions are presented as binary outcomes reflecting presence or absence of toxicity
together with quantitative measures of uncertainty. The approach allows users to identify analogues in different ways,
quickly assess the availability of relevant in vivo data for those analogues and visualize these in a data matrix to
evaluate the consistency and concordance of the available experimental data for those analogues before making a
GenRA prediction. Predictions can be exported into a tab-separated value (TSV) or Excel file for additional review and
analysis (e.g.. doses of analogues associated with production of toxic effects). GenRA offers a new capability of making
reproducible read-across predictions in an easy-to use-interface
iletype T
Md: Filetype ~ ©
-------
^pE!ftip GenRA - Next Steps
Environmental Protection
Agency •
Ongoing research:
Summarising and aggregating the toxicity effect predictions to guide end
users - what effect predictions are we most confident about (digesting A
interpreting the predictions more efficiently)
Consideration of other information to define and refine the analogue
selection & evaluation - e.g. physicochemical similarity, metabolic
similarity, reactivity similarity, bioactivity similarity (transcriptomics
similarity)...
-EPA New Chemical Categories
-Quantifying the impact of physicochemical similarity on read-across
performance (Helman et al., 2018)
National Center for
Computational Toxicology
-------
GenRA - Next Steps
Environmental Protection
Agency
Dose response information to refine scope of prediction beyond binary
outcomes
-Transitioning from qualitative to quantitative predictions - how to apply
and interpret GenRA in screening level hazard assessment
-Starting with quantitative data - e.g. acute rat oral toxicity (Helman et
al (2019), ToxRefOB v2 (Helman et al (2019)
National Center for
Computational Toxicology
-------
Case study: Acute toxicity
Environmental Protection m m
Agency
• Transitioning GenRA to make quantitative predictions
• Investigated extending GenRA using the acute oral rat systemic toxicity data
collected as part of the ICCVAM Acute toxicity workgroup
• NICEATM-NCCT effort to collate a large dataset of acute oral toxicity to
evaluate the performance of existing predictive models and investigate the
feasibility of developing new models
National Center for
Computational Toxicology
-------
United States
Environmental Protection
Agency
Acute toxicity: Dataset creation
Database Resource
Rows of
Data
(number of
LD50
values)
Unique
CAS
ECHA (ChemProp)
5533
2136
JRC AcutoxBase
637
138
NLM HSDB
4082
2238
OECD (eChemPortal)
10206
2314
PAI (NICEATM)
364
293
TEST (NLM ChemlDplus)
13689
13545
Rat oral LD50s:
16,297 chemicals total
34,508 LD50 values
Require unique LD50 values
with mg/kg units
Preprocessing for modelling
Karmaus et al, 2018; Kleinstreuer et al., 2018
National Center for
Computational Toxicology
-------
¦aej* Exploratory Data Analysis
Environmental Protection mm m
Agency
•Found DSSTox matches for 7011 substances
• Extracted MW values
Histogram of LD50 (mgkg)
Histogram of LD5Q (log molar)
0 5000 10000 15000 20000 25000 30000 350 00 40000
LD50 (mgkg)
LD50 (log molar)
National Center for
Computational Toxicology
-------
GenR A approach : Overall global' performance
Environmental Protection | |
Agency
Search for a maximum of 10 nearest neighbours on entire dataset
Use a min similarity threshold of 0.5
Residual Plot
_ra
o
e
en
o
3 -
"S o
"O
(Li
-1 "
-2 -
True vs. Predicted
3 -
2 -
5 1 ¦
¦U
i
oH
-l
-2
-3
Residua
True LD50 {log molar)
National Center for
Computational Toxicology
• Linear regression used to fit predicted and observed LD50
values
• R2 = 0.61
• RMSE = 0.58
• A few outliers, but not too extreme
Residuals clustered around zero with no obvious patterns
46
-------
Environmental Protection
Agency
Coverage vs Similarity vs Performance
Coverage vs Similarity
Coverage vs Similarity
R2 for up to k source analogues
R2 for up to k neighbors
0.4 0.6
Similarity threshold (s
National Center for
Computational Toxicology
Based on the grid searches
performed, k = 10, s = 0.5
were reasonable parameters
to tradeoff coverage vs
prediction accuracy
600 0
Coverage for exactly k neighbors
-------
United States
Environmental Protection
Agency
Monte Carlo Cross Validation
R2 score for 100 75-25 train-test splits
Estimate confidence in R2
75-25 train-test splits
R2 values range from 0.46 to 0.62
GenRA performs robustly on this
acute tox data set
0.46 0 4B 0.50 0.52 0.54 0.56 0.5B 0 60 062 0.64
R2 score
National Center for
Computational Toxicology
Helman et al. (2019)
48
-------
United States
Environmental Protection
Agency
Evaluating 'local' performance
Clustered chemicals into 100
groups on the basis of ToxPrint
fingerprints
Explored performance on the basis of
individual clusters to gauge what sorts
of chemicals resulted in significantly
improved performance (R2) relative to
the overall 'global' performance
reported using 10 nearest neighbours
and a similarity of 0.5
Average R2 values improved
(R2>0.61) for 19 out of the 100
clusters, some
up to 0.91
National Center for
Computational Toxicology
Carbamate containing substances
True vs. Predicted for Cluster 59
0.0 0.5
True LD50 {log molar)
-------
•3.EPA Structure-Activity similarity (SAS) map
Fnvirnnmpntal Prntprtinn • •
United States
Environmental Protection
Agency
Are there pairs of substances that are very similar structurally
with very high LD50 differences, so called activity cliffs
SAS map
Molecular Similarity (ToxPrints)
The number of chemical pairs that
fell within the activity cliff
quadrant was very low relative to
the total number of chemical pairs
captured.
—i
This suggests that the chemical
fingerprints were able to capture
sufficient information to make robust
predictions of acute oral toxicity.
National Center for
Computational Toxicology
-------
SEPA
EPA's Per- and Polyfluoroalkyl
Substances (PFAS) Action Plan
^1 Protection
Using New Approach Methods t<
Fill Information Gaps for PFAS
Research Area 1: What are the human health and ecological effects of
exposure to PFAS?
• Using computational toxicology approaches to fill in gaps. For the many PFAS for which
published peer-reviewed data are not currently available, the EPA plans to use new approaches
such as high throughput and computational approaches to explore different chemical categories
of PFAS, to inform hazard effects characterization, and to promote prioritization of chemicals for
further testing. These data will be useful for filling gaps in understanding the toxicity of those
PFAS with little to no available data. In the near term, the EPA intends to complete assays for a
representative set of 150 PFAS chemicals, load the data into the ComoTox Chemicals Dashboard
for access, and provide peer-reviewed guidance for stakeholders on the use and application of
the information. In the long term, the EPA will continue research on methods for using these
data to support risk assessments using New Approach Methods {NAMs) such as read-across and
transcriptomics, and to make inferences about the toxicity of PFAS mixtures which commonly
occur in real world exposures. The EPA plans to collaborate with NIEHS and universities to lead
the science in this area and work with universities, industry, and other government agencies to
develop the technology and chemical standards needed to conduct this research.
National Center for
Computational Toxicology
~1,223 PFAS currently in TSCA inventory for use in US
~ 602 of those currently active
+ unknown number of degradation and manufacturing
byproducts
EPA 2019 PFAS Action Plan recognised need for approach to
grouping approaches
-------
United States
Environmental Protection
Agency
National Center for
Computational Toxicology
Assembled a PFAS Chemical Library for
Research and Methods Development
• Attempted to procure ~3,000 based on
chemical diversity, Agency priorities, and
other considerations
• Obtained 480 total unique chemicals
• 430/480 soluble in DMSO (90%)
• 54/75 soluble in water (72%)
(incl. only 3 DMSO insolubles)
• Issues with sample stability and volatility
• Categories assigned based on three
approaches
• Buck et al., 2011 categories
• Markush categories
• OECD categories
• Manual assignment
Kathy Coutros, Chris Grulke, arid Ann Richard
PFAS|EPA: ToxCast Chemical Inventory
4j#r
—o
r#°H
—O
4#"
—w
—f
==
-------
United States
Environmental Protection
Agency
Selecting a Subset of PFAS for Tiered
Toxicity and Toxicokinetic Testing
Environmental Health Perspectives
HOME CURRENT ISSUE ARCHIVES COLLECTIONS v v AUTHORS v ABOU
Brief Communication 3 °Pen Access
A Chemical Category-Based Prioritization
Approach for Selecting 75 Per- and
Polyfluoroalkyl Substances (PFAS) for Tiered
Toxicity and Toxicokinetic Testing
Grace Patlewicz, Ann M. Richard, Antony J. Williams, Christopher M. Grulke, Reeder Sams,
Jason Lambert, Pamela D. Noyes, Michael J. DeVito, Ronald N. Hines, MarkStrynar,
Annette Guiseppi-Elie, and Russell S. Thomas
Published: 11 January 2019 CID: 014501 https://doi.org/10.l289/EHP4555
Goals:
• Generate data to support development and
refinement of categories and read-across
evaluation
• Incorporate substances of interest to Agency
• Characterise mechanistic and toxicokinetic
properties of the broader PFAS landscape
Selected 150 PFAS in two phases
representing 83 different categories
• 9 categories with > 3 members
• Lots of singletons
National Center for
Computational Toxicology
Data collection:
Pre-defined
structural
categories
Step 0:
Characterizing the
PFAS library
Maximizing Read-across
A
On Wkgrp-31 list; On EPA-PFAS list;
Availability of in vivo data Availability of in vivo data
Availability of
in vivo data
r
Capturing Structural Diversity
A
EPA interest
in vivo data lacking
Characterizing the
PFAS Landscape
Step 1: Select
substances from
categories of greatest
interest to the Agency
Step 2: Select
substances from
categories of interest to
the Agency
Step 3: Select
substances from
remaining categories
with in vivo data
Step 4: Select
substances from
categories of interest
to the Agency
Step 5: Select
substances from
remaining
categories
Agency interest
5 structural categories
+2 structural categories
+6 structural categories'
*2 categories contained only 1
chemical, so were not included
+5 structural categories
^ 10 structural categories
Availability of in vivo data
53 structural
categories
/s.
53 substances: 12 categories
9 substances:
13 substances
6 categories
10 categories
-------
In Vitro Toxicity and Toxicokinetic Testing
Toxicoloaical Response
Assay
Assay Endpoints
Purpose
Hepatotoxicity
3D HepaRG assay
Cell death and transcriptomics
Measure cell death and changes
in important bioloqical pathways
Developmental Toxicity
Zebraf ish embryo assay
Fertilization, lethality, and
structural defects
Assess potential teratogenicity
Immunotoxicity
Bioseek Diversity Plus
Protein biomarkers across
multiple primary cell types
Measure potential disease and
immune responses
Mitochondrial Toxicity
Mitochondrial membrane
potential and respiration
(HepaRG)
Mitochondrial membrane
potential and oxygen
consumption
Measure mitochondrial health
and function
Developmental
Neurotoxicity
Microelectrode array assay (rat
primary neurons)
Neuronal electrical activity
Impacts on neuron function
Endocrine Disruption
ACEA real-time cell proliferation
assay (T47D)
Cell proliferation
Measure ER activity
General Toxicity
Attagene cis- and trans-
Factorial assay (HepG2)
Nuclear receptor and
transcription factor activation
Activation of key receptors and
transcription factors involved
in hepatotoxicity
High-throughput transcriptomic
assay (multiple cell types)
Cellular mRNA
Measures changes in important
bioloqical pathways
High-throughput phenotypic
profiling (multiple cell types)
Nuclear, endoplasmic reticulum,
nucleoli, golgi, plasma
membrane, cytoskeleton, and
mitochondria morpholoqy
Changes in cellular organelles
and general morphology
Toxicokinetic Parameter
Assay
Assay Endpoints
Purpose
Intrinsic hepatic
clearance
Hepatocyte stability assay
(primary human hepatocytes)
Time course metabolism of
parent chemical
Measure metabolic breakdown
by the liver
Plasma protein binding
Ultracentrifugation assay
Fraction of chemical not bound
to plasma protein
Measure amount of free
chemical in the blood
*Assays being performed by NTP and EPA
National Center for
Computational Toxicology
-------
Current work in progress
Environmental Protection I mm
Agency
How do the structural categories inform read-across? How are
the categories enriched by the bioactivity (tiered toxicity and
toxicokinetic) data being generated?
National Center for
Computational Toxicology
-------
United States
Environmental Protection
Agency
Attagene cis
Assay
Hpa I
Cotransfection J"~L
(transient)
mRNA
RT-PCR and
Hpa I digest
Assay cells
O
E_
o.
o_
o_
¥
A
Capillary
electrophoresis §
TF B
IF A
±
TF C
tt 1 #3
National Center for
Computational Toxicology
and trans- Factorial
• CIS Assay
• 47 Endogenous
Transcription Factors
• Xenobiotic pathways
• Cell
growth/differentiation
• Endocrine pathways
• Stress response
• TRANS Assay
• 24 human nuclear receptors
• GAL-4 formats (NR ligand-
binding domains)
• HepG2 cells
• Concentration-response
testing
• 24-hour exposure
Keith Houck arid Grace Patlewicz
-------
<>EPA Preliminary Category-Based Analysis of the
United States * *
Environmental Protection A » » • » • 1^ _1_ A
Agency Attagene Transcription Factor Assay
Estrogen Receptor Activity
i
U 4.0
3.5
3.0
4.50
=5,4.25
i 4.00
J 3.75
3.50
3.25
3.00
NRF2 Activity
"*r~ ~1
-
T
n
_ , ¦
¦
HtMM »IMM« ^ »•
Mi
PPARa Receptor Activity
£ 3.75
refinedcategories
refinedcategories
refinedcategories
*7 categories with STD > 0.6
National Center for
Computational Toxicology
Keith Houck arid Grace Patlewicz
-------
United States
Environmental Protection
Agency
High-Throughput Phenotypic Profiling
(oka 'Cellular Pathology1)
Multiple Cell
Types
Concentration
Response
Screening
*
Multi-Parameter Cellular
Phenotypic Profiling
DNA RNA/ER AGP Mito H-33342 Casp3/7 PI
Cell Compartments
NUCLEUS RING CYTOPLASM MEMBRANE CELL
<
- y
•
I
Km
r
V
w
oO:
£§» OS?
Ill
Shape (M) Threshold Compactness (C)
o
Radial distribution (R)
*4
Axial (A) Profile (P)
Mode-of-Action
Inference
nydrtKhkwWe Rwerpine
Concentration Response
Modeling
r
?7
Dos*
»0rt»i2—MocfcH
~1,300 endpoints
National Center for
Computational Toxicology
Joshua Harrill arid Johanna Nyffler
-------
United States
Environmental Protection
Agency
Preliminary Category-Based Analysis of the
Phenotypic Profiling Assay
MCF7 Cells
™ 0.5
i/> t/i
w m
aj a;
{/I l/) c/i
u u
ro rn
U (T3
ra n=
m x
a» T3
"D a> in
•= cu
-C l/l
Q_ 03
in _Q
J*
— aj
ra rn
(fi w £
> a)
U20S Cells
C
ro
a;
"a
4.0
3.5
3.0
2.5
2.0
ro
CL
0.0 ' '
IA VI Ul 1/1
~ CD
U» U)
m m tn
in in
QJ QJ
•o ti
QJ QJ QJ
o ia i/i a
ra ra
u ra
ra ra
^ QJ ;»¦
v xj
^ ra ra
— ra ra
ZJ QJ
u 0)
oj ra
oj ra
pairwi5e_eiiclidenn_distanc«
National Center for
Computational Toxicology
Structure ^Category
Joshua Harrill, Johanna Nyffeler, and Grace Patlewicz
-------
Current PFAS Grouping Approaches Use
United States I I I
Environmental Protection # m m - ¦ a m ^ #
Different Levels of Aggregation
c
o
-*—>
03
CD
0
L_
O)
CD
<
CC
L_
=3
O
in
4—
O
0
>
0
*~
1=3 ~ a ~ ~ czi [*] [ *
ABCDEFG H
Chemical Categories/Group
National Center for
Computational Toxicology
*
Available source in vivo tox study
-------
United States
Environmental Protection
Agency
Incorporating Mechanistic and Toxicokinetic
Data to Inform PFAS Category Aggregation
"O
0
E c
o °
c "to
— CD
^ 2
H CD
u jj?
c <
S s
=5
m o
C D
CC J=
-5 w
o
*~
c*
EEC
B C1 C2 D
~
E
r*i
G
H
Chemical Categories/Group
National Center for
Computational Toxicology
* *
Needed in vivo tox study Available source in vivo tox study
-------
v>EPA Challenges with the analysis to date...
Environmental Protection
Agency
• Initially structural category assignments were largely expert
driven
• This was pragmatic based on what resources were available at
the time, however it is difficult to assign membership
reproducibly and objectively with a manual naming convention
• Moreover this does not facilitate profiling of other PFAS
inventories/libraries of interest e.g. OECD
National Center for
Computational Toxicology
-------
United States
Environmental Protection
Agency
PFAS "Categories": Per <& Poly-fluorinated alkyl
substances
Expert"-assigned PFAS categories - manual, subjective
- Buck et al. (DuPont), based on chemical <& series informed by
synthetic pathways (e.g., fluorotelomers)
- data-gathering, occurrence reports, ecotox
- OECD PFAS listing (>4500 chemicals) - manually assigned groupings
National Cente
Computationa
Poly- and Perfluorochemicals
Acyclic - Pure
Atoms: N, P, O, S, Si, CI, Br, I = NOT
# of Cycles = 0
AND
Cyclic - Pure
Atoms: N, P, O, S, Si, CI, Br, I = NOT
BZH # of Cycles > 1
Carboxylic Acids
Atoms: N, P, S, Si, CI, Br, I = NOTl
AND
OjH}
/> CH3
0
Expert category
Fluorotelomer acrylates
Fluorotelomer alcohols
Polyfluorinated alcohols
Fluorotelomer sulfonates
N-alkyl perfluoroalkyl sulfonamidoacetic acids
N-alkyl perfluoroalkyl sulfonamidoethanols
Perfluoroalkyl aldehydes
Perfluoroalkyl amides
Perfluoroalkyl carboxylates
Perfluoroalkyl acyl fluorides
Perfluoro vinyl esters
Perfluoroalkyl ketones
Semi-fluorinated alkenes
Perfluoroalkyl vinyl ethers
Perfluoroalkyl alkyl ethers
Fluorotelomer amines
Perfluoroalkyl sulfonamides
Semi-fluorinated alkanes
Class
Category_Namel
Category_Name2
Alcohol
Fluorotelomer alcohols
Fluorotelomer (linear) n:2 alcohols
Sulfonic Acid
Perfluoroalkyl sulfonic acids
Perfluoroalkyl (linear C4-C10) sulfonic acids
Perfluoroalkyl ethers
Fluorotelomer phosphates
-------
<>epa "Expert-assigned" OECD PFAS Categories,
Environmental Protection
Agency
e.g.
> 4730 PFAS in list
> 173 expert-assigned categories
under 8 general headings (bold)
> Broad "catch-all" terms (in red)
> Structural elements, but NOT
structure-based
> Requires expert to assign new
chemicals to categories
perfluoroalkyl carbonyl compounds
CnF2n+1 C(O) R
perfluoroalkyl carbonyl halides
R = F/CI/Br/l
perfluoroalkyl carboxylic acids (PFCAs),their salts and esters
R = OH, ONa, OCH3, etc.
other perfluoroalkyl carbonyl-based nonpolymers
to be refined
perfluoroalkyl carbonyl amides / amido ethanolsand other alcohols
R = NH2, NH(OH), etc.
perfluoroalkyl carbonyl (meth)acrylate
R = R'_0C(0)CH=CH2
perfluoroalkyl carbonyl (meth)acrylatepolymers
1-H perfluoroalkyl carboxylic acids
H(CF2)nCOOH
perfluoroalkane sulfonyl compounds
CnF2n+1 S(0)(0) R
perfluoroalkane sulfonyl halides
R = F/CI/Br/l
perfluoroalkane sulfonic acids (PFSAs), their salts and esters
R = OH, ONa, OCH3, etc.
perfluoroalkanesulfonyl-based nonpolymers
per- and polyfluoroalkyl ether-based compounds
CnF2n+1 0 CmF2m+1 R
per-and polyfluoroalkyl ether sulfonic acids (PFESAs), their salts
and esters, as well as derivatives
CnF2n+1_0_CmF2m+1_S03H
fluorotelomer-related compounds
perfluoroalkyl iodides (PFAIs)
CnF2n+1_l
n:2 fluorotelomer-based non-polymers
CnF2n+1_C2H4_R, to be refined
National Center for
Computational Toxicology
-------
sepa Translating Expert Categories to Markush
United States
Environmental Protection
Agency
Expert category
Fluorotelom er acrylates
Fluorotelomer alcohols
Polyfluorinated alcohols
Fluorotelomer sulfonates
N-alkyl perfluoroalkyl sulfonamidoacetic acids
N-alkyl perfluoroalkyl sulfonamidoethanols
Perfluoroalkyl aldehydes
Perfluoroalkyl amides
Perfluoroalkyl carboxyiates
Perfluoroalkyl acyl fluorides
Perfluoro vinyl esters
Perfluoroalkyl ketones
Semi-fluorinated alkenes
Perfluoroalkyl vinyl ethers
Perfluoroalkyl alkyl ethers
Fluorotelomer amines
Perfluoroalkyl sulfonamides
Semi-fluorinated alkanes
Perfluoroalkyl sulfonates
Perfluoroalkyl sulfonamido amines
Polyfluoroalkyl carboxyiates
Perfluoroalkyl ethers
Fluorotelomer phosphates
<
<
-------
*»epa Example of Markush representation
Environmental Protection ® ®
Agency
United States
Environmental Protection
l Agency
Advanced Search
Batch Search
Predictions
Downloads
Search All Data
/-m
,„UU
Searched Chemical
-S—OH
I J" II
F 0
Perfluoroalkyl sulfonates
DTXSID: DTXSID70892979
CASRN: NOCAS 892979
markush
Perfluoroheptanesulforjic acid
DTXSID: DTXSID805992Q
CASRN: 375-92-8
markush
o F f F r
OH F F F F
Perfluorobutanesulfonic acid
DTXSID: DTXSID5030030
CASRN: 375-73-5
markush
MO—v==o
II
0
Perfluoropentanesulfonic acid
DTXSID: DTXSID8062600
CASRN: 2706-91-4
markush
Perfiuorooctanesulfonic acid
DTXSID: DTXSID3031864
CASRN: 1763-23-1
markush
Perfluorodecanesulfonic acid
DTXSID: DTXSID3040148
CASRN: 335-77-3
markush
n
o
Perfluorononanesulfonic acid
DTXSID: DTXSID8071356
CASRN: 68259-12-1
markush
F
CJ
=s=
I
OH
markush
OH
II
Perfluorohexanesulfonic acid
DTXSID: DTXSID7040150
CASRN: 355-46-4
Perfluoropropanesulfonic acid
DTXSID: DTXSID30870531
CASRN: 423-41-6
markush
M / .
Perfluorododecanesulfortic acid (PFDOS)
DTXSID: DTXSID20873011
CASRN: 79780-39-5
National Center for
Computational Toxicology
-------
United States
Environmental
Agency
Exploiting fixed fingerprints to facilitate
ProtectioTi ®
objective structural categories
• For the ~150 set,
hove aimed to
harmonise the 3
schemes using fixed
ToxPrints
• Defined rules on
membership based on
specific features
• Extendable to
incorporate other
information i.e.
bioactivity
Final TxP Category Summary
National Center for
Computational Toxicology
Ann Richard and Grace Patlewicz
-------
*»iPA Take home messages
Environmental Protection
Environmental Protection
Agency
* Computational toxicology approaches impact many aspects of regulatory
contexts
• Outlined how computational approaches fit within an I ATA
•Illustrated how we have explored coupling TTC A HTE for a risk-based
prioritization application
* Discussed read-across approaches, tools their frameworks
• Proposed a harmonised framework for read-across approaches
National Center for
Computational Toxicology
-------
*»iPA Take home messages
Environmental Protection
Environmental Protection
Agency
Outlined GenRA, how it was developed and how it is aligned with this
framework - public tool
Initial 6enR A (baseline) considers structural similarity but current work has
evaluated the quantitative impact of physicochemical similarity (as it relates
to bioavailability) and transitioning to dose predictions e.g. acute toxicity
LD50
Highlighted the research efforts of using chemical structural groupings to
underpin selection of representative PFAS for toxicity and toxicokinetic
testing using NAAAs
National Center for
Computational Toxicology
-------
v>EPA
United States
Environmental Protection
Agency
•Many but in particular..
• Imran Shah
• GeorgeHelman
• Tony Williams
• Richard Judson
• Ann Richard
• Chris Grulke
• Keith Houck
• Jason Lambert
• John Wambaugh
• Joshua Harrill
• Johanna Nyffeler
• Rusty Thomas
National Center for
Computational Toxicology
------- |