The Influence of In Vitro Disposition and Toxicokinetics on the Association of In Vitro Bioactivity and In Vivo Toxicity Data

https://orcid.org/0000-0001-7713-985Q
https://orcid.org/0000-0002-4024-534X
United States
Environmental Protection
Agency	.
The Influence of in V Disposition and
Toxicokinetics on the Association of
Bioactivity and in VivoToxicity Data
Gregory Honda1 and John Wambaugh2
1) Office of Air Quality Planning and Standards
U.S. EPA Office of Air and Radiation
2) Center for Computational Toxicology and Exposure
U.S. EPA Office of Research and Development
Society of Toxicology Annual Meeting
New Data and Tools for Understanding Chemical Distribution In Vitro
March 16, 2020
Anaheim, California
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
Office of Research and Development
Center for Computational Toxicology and Exposure
Progress for a Stronger Future

-------
^IPo	Conflict of Interest Statement
United States
Environmental Protection
Agency
The authors declare no conflict of interest
2 of 33
Office of Research and Development

-------
• Research conducted by a combination of Federal
scientists (including uniformed members of the
Public Health Service); contract researchers; and
postdoctoral, graduate student, and post-
baccalaureate trainees
Big! Office of Research and Development
oEPA	US EPA Office of Research and Development
I I + C+o + /-*o
United States
Environmental Protection
Agency
•The Office of Research and Development (ORD) is the scientific research arm of EPA
•562 peer-reviewed journal articles in 2018
• Research is conducted by ORD's four national centers, and three
offices organized to address:
• Public health and en v. assessment; comp. tox. and exposure;
env. measurement and modeling; and en v. solutions and
emergency response.
•13 facilities across the United States
ORD Facility in
Research Triangle Park, NC

-------
vvEPA
United States
Environmental Protection
Agency
Introduction
To use high-throughput screening (HTS) assays as an alternative to traditional
animal studies we must link in vitro bioactivity concentrations and toxic
doses via IVIVE.
Previously, it has not been clear whether the use of IVIVE even improves the
observed association between in vitro bioactivity and in vivo toxicity data.
We have used an in vitro disposition model and a high-
throughput, physiologically based toxicokinetic (PBTK) model
to relate in vitro bioactivity (ToxCast) and endpoint specific rat
in vivo toxicity data.
For every possible comparison of in vitro and in vivo endpoint,
the concordance between the in vivo and in vitro data was
evaluated by a regression analysis.
Dose-Response
(Toxicokinetics
/Toxicodynamics)
Chemical Risk
4 of 33
Office of Research and Development
The NRC (1883) outlined three components for determining chemical risk,

-------
oEPA
/.
United States
Environmental Protection
Agency
High- "hroughput Bioaci /
Projects F
'4
A
•j
)
J\
National Institute of
Environmental Health Sciences
We attempt to estimate points of departure in vitro using high
throughput screening (HTS) for bioactivity as a surrogate for hazard data
Tox21: Examining >8,000 chemicals using ~50 assays intended to
dentify interactions with biological pathways (Schmidt, 2009)
ToxCast (Toxicity Forecast): For a subset (>3000) of Tox21 chemicals
EPA has measured >1100 additional assays-endpoints (Kavlock et
al., 2012)
Most assays conducted in dose-response format (identify 50%
activity concentration - AC50 - and efficacy if data described by a
Hill function, Filer et al., 2016)
All data are public: http://comptox.epa.gov/dashboard/
Office of Research and Development
I To>p7 /
Z-L alvk

©
w
c
o
Q_
(/)
0)
Qd

aantp
Nolttnal bftksbgy ftegrow
NCATS
In vitro Assay AC50
Concentration
CO	CM
CM	CO

-------
oEPA
United States
Environmental Protection
Agency
In Vitro - In Vivo Extrapolation (IVIVE)
What do we do with an in vitro concentration? - IVIVE is the use of in vitro experimental data to predict
phenomena in vivo
• IVIVE-PK/TK (Pharmacokinetics/Toxicokinetics):
•	Fate of molecules/chemicals in body
•	Considers absorption, distribution, metabolism, excretion (ADME)
•	Uses empirical PK and physiologically-based (PBPK) modeling
IVIVE-PD/TD (Pharmacodynamics/Toxicodynamics):
•	Effect of molecules/chemicals at
biological target in vivo
•	Assay design/selection important
•	Perturbation as adverse/therapeutic
effect, reversible/ irreversible effeccts
Both contribute to in vivo effect prediction
Office of Research and Development
Normalization of dose
PBPK models
NRC (1998)
Humans: in vivo
Testable predictions
Extrapolation
using PD and
PBPK models
Comparative testing

-------
oEPA
United States
Environmental Protection
Agency
1000
100
I
4
«
u
E
o
u
e
c
0.01
0.001
Comparing on the Basis of Concentration
1000
100 3
u
e
o
u
n
£
U1
ra
10
0.1
0.01
0.001
o
E
3

5
Triclosan MBP MEHP PFOA 2,4-D
(90/615) (8/615) (35/615) (24/615) (10/615)
Range of bioactive concentrations
across ToxCast assays
^ Estimated or measured
average concentrations
associated with the LOAEL
in animal studies
O NOAEL in animal studies
£ Humans with chronic
exposure reference values
(solid circles)
X Volunteers using products
containing the chemical
+ Bio-monitored occupational
populations
A General populations
The five chemicals (as of 2011) with plasma biomonitoring AND ToxCast data... what do we do about the other 1000's?
Office of Research and Development
Ay I ward and Hays (2011)

-------
oEPA
United States
Environmental Protection
Agency
High Throughput Toxicokinetics (HTTK)
Most chemicals lack public toxicokinetic-related data (Wetmore et al., 2012):
In vitro toxicokinetic data + generic toxicokinetic model
= high(er) throughput toxicokinetics
^ v # v ^ v #
p l
a.
i 2^ ¦=> ..¦=>
Primary
Compartment
si,
Oral Absorption
httk
Office of Research and Development
Metabolism
Renal Clearance

-------
oEPA
United States
Environmental Protection
Agency
Comparing IVIVE Predictions with
Toxic Doses
Rat-specific HTTK data were collected in vitro for ~50 chemicals, allowing IVIVE with ToxCast Data
3
Q)
5
d 2
o
£
LU _
1
DO ®
3 E
0)
x £
o O
I-
ft 10
>*
> o
c
0
£ -2
E
3
E
'c
-3
-4
•
•
A
•
A A ^ /
•
* y
• y
•• X
a /
•« *.
% •
• • • • X • •
•
•


5.7% below line
-4-3-2-10	1	2	3
Minimum In Vitro Rat Oral Equivalent Dose (mg/kg/d)
Office of Research and Development

-15
-10
-5
10"2 10"1 10° 101 102 103 104 105
Log10 Ratio ToxRefDB Min LEL:ToxCast
Min Oral Equivalent Dose
Distribution Summary Statistics
Median	1.82 (66.07)
Upper Quartile 2.55 (354.81)
Lower Quartile 0.95 (8.91)
Wetmore et al. (2013)

-------
vvEPA
United States
Environmental Protection
Agency
There Are Many Considerations When
Doing IVIVE
in vitro
(nominal testing concentration)
~ l Media/Air
Chemica
'6
[C,
nominalJ
o
o
Plastic [^free,invitrol~^up[^ nominal]
Media
$ Lipid
^ and
Binding
Cell Binding
Protein
Binding
[Cone.] In Vitro
in vivo
(mg/kg bodyweight/day)
Plasma
Tissue
[C,
t 30dl
> [^plas mal
[^bloodl/^b:p
[^free, plasma!
^up[^plas mal
[^tissue]
^p[^free,plasma]
Renal Clearance
f *n *rr	l
up ^GFR L kidney,plasmaJ
Restrictive Metabolic Clearance
[C UVer,plasma\
liver,plasma\
Qliver * fup *
Qliver "I" fup * [^i
OR Non-Restrictive Metabolic Clearance
Qliver * \pliver,plasma]
Qliver "I" [Qtver,p/asma]
How do you select the appropriate in vitro and in vivo concentrations for extrapolation?
10 of 33
Office of Research and Development

-------
oEPA
United States
Environmental Protection
Agency

Inhaled Gas

Lung Tissue

	~
Lung Blood
	~

Venous Blood
Qgfr
<4	

Qkidney
*	
Arterial Blood
Kidney Tissue
Kidney Blood




Gut Lumen
Qgut
<	
Gut Blood






t
Qmetab
Liver Tissue
QgL
Liver Blood




Qliver
-4	
Rest of Body
Qrest
<*	
Body Blood
Pearce et al. (2017)
Office of Research and Development
A General Physiologically-based
Toxicokinetic (PBTK) Model
•	R package "httk" includes a generic PBTK model
•	Can be tailored to a chemical using in vitro data and predictions from
chemical structure
•	Some tissues (e.g. arterial blood) are simple compartments, while others
(e.g. kidney) are compound compartments consisting of separate blood and
tissue sections with constant partitioning (i.e., tissue specific partition
coefficients)
•	Some specific tissues (lung, kidney, gut, and liver) are modeled explicitly,
others (e.g. fat, brain, bones) are lumped into the ''Rest of Body"
compartment.
•	The only ways chemicals "leave" the body are through metabolism (change
into a metabolite) in the liver or excretion by glomerular filtration into the
proximal tubules of the kidney (which filter into the lumen of the kidney).

-------
vi-EPA
United States
Environmental Protection
Agency
• In vivo data for rat were accessed from the
Toxicity Reference (ToxRef) database version 1
• Much of the data in ToxRefDB vl was derived
from studies or study summaries for study
designs compliant with or similar to the EPA
OCSPP 870 series guidelines
• ToxRef DB vl is a "positives-only" database, and in
vivo data were reported as the nominal dose at
which an effect (not necessarily critical) was
observed for a particular endpoint
• The analysis in this work included chronic (2
year), subchronic (90 day), and developmental
(parental and fetal generations) study types
12 of 33
Office of Research and Development
ToxRefDB
Research
volume 1171 kl/mber i | March 2009 * Environmental Health Perspectives
Profiling Chemicals Based on Chronic Toxicity Results from the U.S. EPA
ToxRef Database
Matthew T. Martin, Richard S. Judson. David M. Rati, Robert J. Kavlock, and David J. Dix
National Canter for Computational Toxicology. Office of Research and Development. U.S. Environmental Protection Agency. Re»earch
Triangle Park. North Carolina. USA
Bat X(.RiU M): lkirt> ttan of pesticide registration toxicity diu have been historically »torc«J as
hardcopy and tuaiwd documents by tike U.S. Environmental Protection Agency (EPA). A ugnifi
cant portion of these data have nam been processed into itindtnlual and structured toxicity data
within the fcPA's toxicity Reference Database (ToxRefDB). including chronic, cancer, develop
mental, and repeoducuve studies from laboratory animak. Ihese data are now accessible and mine-
able wittun IoxRefDB and arc serving as a primary source of validation for U-S. EPA'i ToxCast
research program in prcd* tiv* toxicology.
OB|K TIVE&x We profiled im mm toxkilin across 310 chemicals as a model application of
ToxRefDB. meeting the need foe detailed anchoring end points for development of ToxCast predic-
tive og natures
Ml 1 HODS: Using query and structured data-mining approaches, we generated toxicity profiles from
ToxRefDB bated on long-term rodent hnnuyi These chronic/cancer data were analyzed for suit-
ability as anchoring end poants based on incidence, target organ, seventy. potency , and ugnificance.
Rim. LI v Under conditions of the bsuassayv we observed pathologies for 273 of 310 chemicals, with
greater preponderance (> 90%) occurring m the liver, kidney, thyroid, lung, testis, and spleen. Wc
observed proliferative lesions for 22S chemicals, and 16" chemicals caused progression to cancer-
rc-latc-d pathologies
Conclusions: Based on incidence, severity , and potency , we selected 26 primarily tissue-specific
pathology end points to umformb classify the 310 chemicals. The resulting toxicity profile classifi-
cations demonstrate 'I** utiiitv of structuring legacv toxicity informal ion and facilitating the com-
putation of these data within ToxRefDB far ToxCast and other applications.
Key VOIDS: cancer, chrome toxicity. pesticides, relational database, toxicity profile. Enrtron Health
Perfect 117.392-399 (2009). doc 10.1289/ehp.08000"4 available via bapJldx.4ai.ffl [Online
20 October 2008]
set of toxicologic information. Ihc complete
and highly standardized data set provided by
ToxRefDB facilitates analysis of the ToxCast
phase 1 chemicals across chemical, study type,
species, target organ, and effect. Additionally.
ToxRefDB serves as a model for other efforts
to capture quantitative, tabular toxicology
data from legacy and new studies and to make
these data useful for cross-chemical computa-
tional toxicology analysis.
Methods
Data characteristics. Wc collected reviews
of registrant-submitted toxicity studies,
known as data evaluation records (DERs), tor
roughly 400 chemicals from the U.S. EI*A s
Office of Pesticide Programs (OPP) within
the Office of Pollution Prevention and I oxic
Substances (OPPTS). The file types of the
DERs include TIFF, Microsoft Word, Word
Perfect, and PDF formats, some of which arc
not directly text-readable. Wc indexed every
DER file based on a file name convention
that consisted of the pesticide chemical (PC)
code, study identification number (MR1D),
«»uly rypr ulrnnhrjunn nnmhrr ihiirri nn
Martin et al. (2009)

-------
vi-EPA
United States
Environmental Protection
Agency
• New rat-specific HTTK data collected for ~80
chemicals in addition to ~50 from Wetmore
et al. (2013)
• For each ToxRef endpoint (mg/kg/day) we
did a forward dosimetry calculation
(predicted \iM concentration)
• For each ToxCast endpoint (|iM) we did a
reverse dosimetry I VIVE calculation
(predicted mg/kg/day dose)
• We compared each ToxRef and ToxCast
endpoint on both |iM and mg/kg/day scales
13 of 33
Calculated Orthogonal Root Mean Squared
Error (ORMSE) - lower is better
Office of Research and Development
Honda et al. (2019)
Example for a single ToxCast assay and ToxRef
endpoint, each point is a chemical:
o
o
<
O)
O
TJ
a>
N
03
~o
i5
to
c) x = Dose
' ORMSE: 0.849
0
2
Honda et al. (2019)

-------
vvEPA
United States
Environmental Protection
Agency
• New rat-specific HTTK data collected for ~80
chemicals in addition to ~50 from Wetmore
et al. (2013)
• For each ToxRef endpoint (mg/kg/day) we
did a forward dosimetry calculation
(predicted |iM concentration)
• For each ToxCast endpoint (|iM) we did a
reverse dosimetry IVIVE calculation
(predicted mg/kg/day dose)
• We compared each ToxRef and ToxCast
endpoint on both |iM and mg/kg/day scales
14 of 33
Calculated Orthogonal Root Mean Squared
Error (ORMSE) - lower is better
Office of Research and Development
Honda et al. (2019)
o
LO
o
<
CD
o
T3
N
T3
CO
TJ
£Z
CO
¦M
00
Example for a single ToxCast assay and ToxRef
endpoint, each point is a chemical:
c) x = Dose



~




•


y
t
•
*
'' ORMSE: 0.849
Honda et al. (2019)

-------
vi-EPA
United States
Environmental Protection
Agency
• New rat-specific HTTK data collected for ~80
chemicals in addition to ~50 from Wetmore
et al. (2013)
• For each ToxRef endpoint (mg/kg/day) we
did a forward dosimetry calculation
(predicted \iM concentration)
• For each ToxCast endpoint (|iM) we did a
reverse dosimetry I VIVE calculation
(predicted mg/kg/day dose)
• We compared each ToxRef and ToxCast
endpoint on both |iM and mg/kg/day scales
15 of 33
Calculated Orthogonal Root Mean Squared
Error (ORMSE) - lower is better
Office of Research and Development
Honda et al. (2019)
Example for a single ToxCast assay and ToxRef
endpoint, each point is a chemical:
c) x = Dose
LD
O
<
D)
O
"O
cu
N
TJ
l_
CD
"O
cz
CD
-»—»
09

/ ORMSE: 0.849
-I	1	T	»
-4 -2 0 2
ible
a)
to
o
"D
cn
"D
aj
N
CD
"D
c
CD
cn
f) x = AC
50
.Af
'r
ORMSE: 0.849
—i	1	i	r-
-4-2 0 2
ihlei
Honda et al. (2019)

-------
vi-EPA
United States
Environmental Protection
Agency
• New rat-specific HTTK data collected for ~80
chemicals in addition to ~50 from Wetmore
et al. (2013)
• For each ToxRef endpoint (mg/kg/day) we
did a forward dosimetry calculation
(predicted \iM concentration)
• For each ToxCast endpoint (|iM) we did a
reverse dosimetry I VIVE calculation
(predicted mg/kg/day dose)
• We compared each ToxRef and ToxCast
endpoint on both |iM and mg/kg/day scales
• Calculated Orthogonal Root Mean Squared
Error (ORMSE) - lower is better
0 Office of Research and Development
Honda et al. (2019)
We compared the ORMSE for dose vs. AC50 with
using PBTK to perform IVIVE:
a) x = C
PBTK
o
LO
O
<
o
o>
o
T3

«
o
"D
o
6)
"O
cl)
N
CO
¦O
c
nj
¦+—'
00




/
/


J
P

8
A

/ ORMSE: 0.89
~
t—

m
-2
f) x = AC
50
"jr
o, *
-------
vvEPA
United States
Environmental Protection
Agency
•	>1000 ToxCast assay endpoints
•	106 specific ToxRef endpoints (68
pathological responses and 3 study types)
•	80 chemicals with observed effects in
ToxRef and bioactivity in ToxCast
17 of 33
Office of Research and Development
"Significance"
-------
vi-EPA
United States
Environmental Protection
Agency
•	>1000 ToxCast assay endpoints
•	106 specific ToxRef endpoints (68
pathological responses and 3 study types)
•	80 chemicals with observed effects in
ToxRef and bioactivity in ToxCast
JEllY 8EPNS
Cause acne::
SCI&JTT5T5!
INVESTIGATES
Borate:
(WlNG
nwftwipr!
F(NE,
WL RXJrJDNO
UNK QDUEEM
JELLY BEANS WD
ACNE (P > 0,05"}.
THAT SEHIES THAT
I HEAR ITS ONLY
A CERTAIN COLOR
THflTCAUSES it:
SCIENTISTS!
M'lhwECftftFT!

18 of 33
Office of Research and Development
'Significance^
https://xkcd.com/882/
-------
oEPA
United States
Environmental Protection
Agency
>1000 ToxCast assay endpoints
106 specific ToxRef endpoints (68
pathological responses and 3 study types)
80 chemicals with observed effects in
ToxRef and bioactivity in ToxCast
JEllY 8EPNS
Cause acne::
SCI&JTT5T5!
INVESTIGATES
(WlNG
nwftwipr!
fine.
VL RXJrJDNO
UNK QDUEEM
JELLY BEAMS WD
ACNE (P > 0,05"}.
That SETTIES THAT
I HEAR ITS ONLY
A CERTainJ Color
TWQciuses it:
SCIENTISTS!
PUT
MWlwECftftFT!
w
Office of Research and Development
WER3ONDN0
UNK BE3VEEN
ft«Pl£ Kay
BOWS AND ACNE
( P > 0.O5J
)
WEftONDNO
UNK, BE3WEEN
5ALMOM KaY
BEANS AND AOIE
( p>0,OT).
)
WERDUNDW
UNKQOUEEM
GREYKUY
BEAMS AND ACNE
( p > 0 .C5),
/
WERXJNDNO
UNK BOWEEW
Beige KaY
Beams and aqc
(p > 0®)
1
a
Significance
ff
WERXJWDNO
UNKGDUEEM
BROW* JEuy
BEANS AND ACNE
( p > 0.O5),
/
WE RX1NDN0
UNKBEWEEN
redkuy
BEANS ANO ACME
(p>O.Oj).
I
WE FOUND NO
UNK (3EJVIEEN
TAn! 3&iy
BEAMS WO ACME
( p > 0.O5).
/
WERXJNDNQ
UNK BETWEEN
UlAC KUY
BEAMS*© ACNE
(P>OOf)
/
WER)W4ON0
UNK BETWEEN
PIN»tKaY
BEANS PND ACNE
(pXXO?).
I
WE. R)UNDN0
UNK BETWEEN
TURCOOI5C KaY
Beans and Acme
(p >0.059.
)
WEI FOUND NO
UNKCOWEEM
CYAN KaY
BEANS PNDACNE
(p>o,o?},
I
WE. FOUND NO
UNK BETWEEN'
BWH KO.Y
Beans wd acne
(p>0.O5).
/
WERXtNDNO
UNK BETWEEN
blue Kay
BEAMS AND ACNE
(p>0.05).
/
WE FOUND NO
UNK BE3UEEN
kiag&jia Kay
BEAMS AND ACNE
(p > 0.0?),
/
WEI FOUND ft
UNKQOUEEM
green Kay
BEAMS AND ACNE
(p O&ffj
)
WEKWNDNO
unkbqueem
teal Kay
BEAMS AND ACNE
(P > 0.05^,
)
WERXMDNO
UNK BE3UEEN
muw Kuy
BEAMS AND ACNE
(p>0 .OS).
>
WE FOUND NO
UNK 0E3UEEN
rwv£ Kuy
beams and acne
{p > o.os),
1
WE FOUND NO
UNKBOWEEH
oeMGEKay
BEAMS AND ACNE
(p>0O5)
/
https://xkcd.com/882/
-------
oEPA
United States
Environmental Protection
Agency
>1000 ToxCast assay endpoints
106 specific ToxRef endpoints (68
pathological responses and 3 study types)
80 chemicals with observed effects in
ToxRef and bioactivity in ToxCast
JEllY 8EPNS
Cause acne::
SCI&JTT5T5!
INVESTIGATES
Gut^Ise
(WlNG
nwftwipr!
fine.
VL RXJrJDNO
UNK QDUEEM
JELLY BEAMS WD
ACNE (P > 0,05"}.
That SETTIES THAT
I HEAR ITS ONLY
A CERTainJ Color
TWQciuses it:
SCIENTISTS!
PUT
MWlwECftftFT!
w
Office of Research and Development
WER3ONDN0
UNK BE3VEEN
ft«Pl£ Kay
BOWS AND ACNE
( P > 0.O5J
)
WEftONDNO
UNK, BE3WEEN
5ALMOM KaY
BEANS AND AOIE
( p>0,OT).
)
WERDUNDW
UNKQOUEEM
GREYKUY
BEAMS AND ACNE
( p > 0 .C5),
/
WERXJNDNO
UNK BOWEEW
Beige KaY
Beams and aqc
(p > 0®)
1
a
Significance
ff
WERXJWDNO
UNKGDUEEM
BROW* JEuy
BEANS AND ACNE
( p > 0.O5),
/
WE RX1NDN0
UNKBEWEEN
redkuy
BEANS ANO ACME
(p>O.Oj).
I
WE FOUND NO
UNK (3EJVIEEN
TAn! 3&iy
BEAMS WO ACME
( p > 0.O5).
/
WERXJNDNQ
UNK BETWEEN
UlAC KUY
BEAMS*© ACNE
(P>OOf)
/
WER)W4ON0
UNK BETWEEN
PIN»tKaY
BEANS PND ACNE
(pXXO?).
I
WE. R)UNDN0
UNK BETWEEN
TURCOOI5C KaY
Beans and Acme
(p >0.059.
)
WEI FOUND NO
UNKCOWEEM
CYAN KaY
BEANS PNDACNE
(p>o,o?},
I
WE. FOUND NO
UNK BETWEEN'
BWH KO.Y
Beans wd acne
(p>0.O5).
/
WERXtNDNO
UNK BETWEEN
blue Kay
BEAMS AND ACNE
(p>0.05).
/
WE FOUND NO
UNK BE3UEEN
kiag&jia Kay
BEAMS AND ACNE
(p > 0.0?),
/
WEI FOUND ft
UNKQOUEEM
green Kay
BEAMS AND ACNE
(p O&ffj
)
WEKWNDNO
unkbqueem
teal Kay
BEAMS AND ACNE
(P > 0.05^,
)
WERXMDNO
UNK BE3UEEN
muw Kuy
BEAMS AND ACNE
(p>0 .OS).
>
WE FOUND NO
UNK 0E3UEEN
rwv£ Kuy
beams and acne
{p > o.os),
1
WE FOUND NO
UNKBOWEEH
oeMGEKay
BEAMS AND ACNE
(p>0O5)
/
Neit} s
GREEN OEUY
BEANS DNKED
to acne!
my 5% CNflMtr
CONOODJCEi
SOFNTTiSre-
https://xkcd.com/882/
-------
vi-EPA
United States
Environmental Protection
Agency
• New rat-specific HTTK data collected for ~80
chemicals in addition to ~50 from Wetmore
et al. (2013)
• For each ToxRef endpoint (mg/kg/day) we
did a forward dosimetry calculation
(predicted \iM concentration)
• For each ToxCast endpoint (|iM) we did a
reverse dosimetry I VIVE calculation
(predicted mg/kg/day dose)
• We compared each ToxRef and ToxCast
endpoint on both |iM and mg/kg/day scales
• Calculated Orthogonal Root Mean Squared
Error (ORMSE) - lower is better
0 Office of Research and Development
Honda et al. (2019)
As a sanity check, we also performed IVIVE using
PBTK for a randomly selected chemical:
a) x = C
PBTK
b) x Crancionri
c) x = Dose
O
LO
o
<
o
CT>
o
ID
CD
N
T3
flj
"O
cz
ro
¦M
09


/
/






\
\
N
1	


- ORMSE: 0.812
t. . . . -


~

JfL
•




/ o
~


'' ORMSE: 0.863
s



/
~




•



~
~
~
m

t
'' ORMSE: 0.849
-2
2 -4 -2 0 2 -4 -2
Standardized log10 x-variable
a>
° ?
"O 2
o
6>
o 0
XJ
<1>
£ -2
2
ro
"D
CD
¦+-*
09
d) x AEDpbjk e) x AEDranCj0m
f) x = AC







J


8


•/
/ ORMSE: 0.89
£	

•

ORMSE: 0.809
50

ORMSE: 0.849
2 -4 -2 02 -4-2 02
Standardized log10 x-variable
Honda et al. (2019)
-------
vvEPA
United States
Environmental Protection
Agency
500-
400-
300-
200-
100-
Distribution of ORMSE
For each in
vitro-in vivo
endpoint
pair, we
calculate
the ORMSE
across all
available
chemicals
c
3 0
O
a 500-
400-
300 -
200-
100-
0 --
Dose or AC
50
~n
tu
o
<

to
O
Lower values
indicate lesser error
)
22 of 33
Office of Research and Development Number of Chemicals 5:10 II 10.20 I - 20	Honda et ol. (2019)
-------
vvEPA
United States
Environmental Protection
Agency
500-
400-
300-
200-
100-
For each in
vitro-in vivo
endpoint
pair, we
calculate
the ORMSE
across all
available
chemicals
c
3 0
O
a 500-
400-
300 -
200-
100-
o --
Distribution of ORMSE
Random
Dose or AC
50
"T|
O
£U
a
o
<
-------
vvEPA
United States
Environmental Protection
Agency
500-
400-
300-
For each in
vitro-in vivo
endpoint
pair, we
calculate
the ORMSE
across all
available
chemicals
<+•>
c

3
0-
o

o
500-

400-

300 -

200-

100-
Distribution of ORMSE
PBTK
Random
Dose or AC
50
0.5	1 o
ORMSE
Lower values
indicate lesser error
Randomly selecting
the chemical for the
IVIVE increases
error (on average)
Using PBTK lowers
the error
24 of 33
Office of Research and Development Numbsf of Chsmicsls 5-10 I 10.20 I — 20
Honda et ol. (2019)
-------
vvEPA
United States
Environmental Protection
Agency
What About In Vitro Distribution?
(Please	Stop Discu
Armitage et al. (2014) suggest that in vitro
partitioning relates strongly to logKow and
serum in the medium
Sorption to plastic played a smaller role in
determining the cellular concentration
We can check to see if using an in vitro
disposition model improves IVIVE (that is,
reduces error in comparisons between in
vivo and in vitro endpoints)
Note, Armitage model expanded to
ionizable compounds by Fischer et al.
(2017)
Mass-balance model:
DMSO (dimethyl sulfoxide, a typical
solvent), OM (organic matter)
DMSO
(if present)
N

"7
Bfr
¦•Br,
Sorption to
vessel wall

\
\
W
w
Serum constituents
(if present)
Dissolved
v OM J
Test medium
Cells/tissue
25 of 33
Office of Research and Development
Armitage et al. (2014)
-------
oEPA
United States
Environmental Protection
Agency
Impact of IVIVE Assumptions
Different combinations of assumptions,
for example:
res-tot-vein-mean = restrictive
metabolism, total chemical, venous
concentrations, mean concentration
during tox study
PBTK Random Dose
Office of Research and Development
£
m
OJ
E
¦
C
t
i
Zl
£
it
"d
o >
T 3
$
-------
vvEPA
United States
Environmental Protection
Agency
Impact of IVIVE Assumptions
Different combinations of assumptions,
for example:
res-tot-vein-mean = restrictive metabolism,
total chemical, venous concentrations, mean
concentration during tox study

40 ¦

3D-
o

T3

o

£
20-


(fl

o

m
10-
«o

03

T3

QJ

O

ji

0)

CA

(A

O

E

i-

<•-

o



c

3

o

o

c
(5
03
f 1
E ©
flj ~u
> <0
w
qj
27 of 33
¦ petk | Random Dose
Office of Research and Development
-------
oEPA
United States
Environmental Protection
Agency
Impact of IVIVE Assumptions
Different combinations of assumptions,
for example:
res-tot-vein-mean = restrictive metabolism,
total chemical, venous concentrations, mean
concentration during tox study
TJ
O
40
3D
20
W
at
m 10
t/i
m
U o
o
JV
0)
O)
* 15
Of
E
i- 10
c
3
o
o
08
03	S*
?	3
E	*
qj}	tj
*	?
"5	>
T	i
W	-C
0	—
PBTK I Rsndon
Office of Research and Development
-------
oEPA
United States
Environmental Protection
Agency
Impact of IVIVE Assumptions
Different combinations of assumptions,
for example:
res-tot-vein-mean = restrictive metabolism,
total chemical, venous concentrations, mean
concentration during tox study
TJ
O
40
3D
20
W
at
m 10
t/i
m
U o
o
JV
0)
O)
* 15
Of
E
i- 10
c
3
o
o
08
03	S*
?	3
E	*
qj}	tj
*	?
"5	>
T	i
W	-C
0	—
PBTK I Rsndon
Office of Research and Development
-------
oEPA
United States
Environmental Protection
Agency
Impact of IVIVE Assumptions
Different combinations of assumptions,
for example:
res-tot-vein-mean = restrictive metabolism,
total chemical, venous concentrations, mean
concentration during tox study
nres-tot-tis-max = non-restrictive
metabolism, total chemical, tissue
concentrations, max conc. during tox
study
TJ
O
40
3D
20
W
at
ca 10
0)
as
U o
o
JV
0)
if)
V)
0)
E
15
10
c
3 5
O
O
T]
E
c	«
£ S	E
?li
E	<3i
"a	>
I?
O >	T
w C	Ji*
fl> v	h-
c
03
CD
<10
>:
to
£
vj

£
a
E
9

PBTK I Random Dose
K
a>
E
I
c
a>
>
i
v
eft


J
a>

d;
¦
V
43
Office of Research and Development
-------
oEPA
United States
Environmental Protection
Agency
Impact of IVIVE Assumptions
Different combinations of assumptions,
for example:
res-tot-vein-mean = restrictive metabolism,
total chemical, venous concentrations, mean
concentration during tox study
nres-tot-tis-max = non-restrictive metabolism,
total chemical, tissue concentrations, max conc.
during tox study
Several I VIVE combinations using the
Armitage model decreased error, but no
single ideal approach
at
-a
o
40
3D
20
a>
CO 10
re
o
a>
itl
* 15
tit
E
10
c
3
o
o
c	«
S s	E
?li
E	<3i
"a	>
I?
O >	T
£§	a*
w -C.
P


c
03
CD
«L0
Q>
to
£
vi
¦03
£
a
E
9

IS
05
E
I
c
a>
>
i
v
eft
PBTK I Random Dose

ra
1
<
i
c
<0
4D
E
¦
c
'53
>
dj
a>
ra
c
ra
HI
E
.
c
f
03
ID
C
05
0*
E
?
v>
Q}
*
e
V
c


J
a>
&

dj
¦
$

E
J
a>
a>
Office of Research and Development
V)
P
03
W
qp
9>
-------
oEPA
United States
Environmental Protection
Agency
Comparing Points of Departure and IVIVE
©
>
®
d 2
o
CD
S3 1
o|>
CQ m „
D E o
H— *-
0 >,
DC ¦§*
X >5
o O -i
J— +-
n
got
>1-2
E
3
E
'E
-3
-4
Wetmore et al. (2013)
•
•
A * A /

f 9 y
• r
•• x
A /
1 •
• # /
• •• •• •
• /
* •
•
•


5.7% below line
-4	-3-2-10	1	2
Minimum In Vitro Rat Oral Equivalent Dose (mg/kg/d)
Honda et al. (2019)
a) restrictive, mean
total plasma conc.
4-
o
g
CL
° OH
o>
_o
-2
RMSE: 1.96
ORMSE: 0.918
-3 0 3 6
log10AED10
c) restrictive, mean free
plasma conc., Armitage
4-
§ «
CL
5 OH
ro
_o
-2-

RMSE: 1.4
ORMSE: 0.931
Office of Research and Development
-3 0 3
I°9ioAED10
b) restrictive, mean
free plasma conc.
4-
§ «
CL
£ 0
O
4-
§ «
CL
S 0
Hi
O
-2-
	/
/

-m:

m

•• • ••
'RMSE: 1.68
/ ORMSE: 0.991

1 1 1
3 0 3 6
log10AED10
d) nonrestrictive,
mean tissue conc.

/
•

•

• f' # -

RMSE: 1.82

ORMSE: 0.977
-3 0 3
log10AED10
-------
vvEPA
United States
Environmental Protection
Agency
Summary
•	NAMs for TK allow risk-based prioritization of large numbers of chemicals
•	In vitro disposition modeling and PBTK enable improved via in vitro-in vivo
extrapolation (IVIVE)
•	We tested various sets of IVIVE assumptions and demonstrate
that the combination of PBTK and in vitro disposition modeling
improves our ability to observe the association between in vitro
bioactivity and in vivo toxicity data.
•	Potency values from in vitro screening should be transformed
IVIVE to build better machine learning and other statistical
models for predicting in vivo toxicity in humans
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
Chemical Risk
Dose-Response
(Toxicokinetics
/Toxicodynamics)
33 of 33
Office of Research and Development
-------
Acknowledgements
•	Robert G. Pearce
•	Ly Ly Pham
•	R. Woody Setzer
•	Barbara A. Wetmore
•	Nisha S. Sipes (NTP)
•	Jon Gilbert (Cyprotex)
•	Briana Franz (Cyprotex)
•	Russell S. Thomas
-------
References
v>EPA
United States
Environmental Protection
Agency
Cohen Hubal, EA, et al. "Advancing internal exposure and
physiologically-based toxicokinetic modeling for 21st-
century risk assessments." Journal of exposure science &
environmental epidemiology (2018).
Eissing, Thomas, et al. "A computational systems biology
software platform for multiscale modeling and simulation:
integrating whole-body physiology, disease biology, and
molecular reaction networks." Frontiers in physiology 2
(2011): 4.
Frank, Christopher L., et al. "Defining toxicological tipping
points in neuronal network development." Toxicology and
applied pharmacology 354 (2018): 81-93.
Honda, Gregory S., et al. "Using the concordance of in vitro
and in vivo data to evaluate extrapolation assumptions."
PloS one 14.5 (2019): e0217564.
Jamei, et al. "The Simcyp® population-based ADME
simulator." Expert opinion on drug metabolism & toxicology
2009b;5:211-223
Jongeneelen, Frans J., and Wil F. Ten Berge. "A generic,
cross-chemical predictive PBTK model with multiple entry
routes running as application in MS Excel; design of the
model and comparison of predictions with experimental
results." Annals of occupational hygiene 55.8 (2011): 841-
864.
Office of Research and Development
Lukacova, Viera, Walters. Woltosz, and Michael B. Bolger.
"Prediction of modified release pharmacokinetics and
pharmacodynamics from in vitro, immediate release, and
intravenous data." The AAPS journal 11.2 (2009): 323-334.
McLanahan, Eva D., et al. "Physiologically based
pharmacokinetic model use in risk assessment—why being
published is not enough." Toxicological Sciences 126.1
(2011): 5-15.
National Research Council. (1983). Risk Assessment in the
Federal Government: Managing the Process Working
Papers. National Academies Press.
National Research Council. (2007). Toxicity testing in the
21st century: a vision and a strategy. National Academies
Press.
Obama, B. H. "Executive Order 13642: Making Open and
Machine Readable the New Default for Government
Information. Washington, DC: Office of the Executive."
(2013).
Ring, Caroline L., et al. "Identifying populations sensitive to
environmental chemicals by simulating toxicokinetic
variability." Environment International 106 (2017): 105-118.
Rotroff, Daniel M., et al. "Incorporating human dosimetry
and exposure into high-throughput in vitro toxicity
screening." Toxicological Sciences 117.2 (2010): 348-358.
Sayre, Risa et al., "Database of pharmacokinetic time-series
data and parameters for XX environmental chemicals" in
preparation
Sipes, Nisha S., et al. "An intuitive approach for predicting
potential human health risk with the Tox21 10k library."
Environmental science & technology 51.18 (2017): 10786-
10796.
Wambaugh, John F., et al. "Evaluating in vitro-in vivo
extrapolation of toxicokinetics." Toxicological Sciences
163.1 (2018): 152-169.
Wambaugh, John F., et al. "Assessing Toxicokinetic
Uncertainty and Variability in Risk Prioritization"
Toxicological Sciences (2019), in press
Wang, Ying-Hong. "Confidence assessment of the Simcyp
time-based approach and a static mathematical model in
predicting clinical drug-drug interactions for mechanism-
based CYP3A inhibitors." Drug Metabolism and Disposition
38.7 (2010): 1094-1104.
Wetmore, Barbara A., et al. "Integration of dosimetry,
exposure and high-throughput screening data in chemical
toxicity assessment." Tox. Sciences (2012)
Wetmore, Barbara A., et al. "Incorporating high-throughput
exposure predictions with dosimetry-adjusted in vitro
bioactivity to inform chemical toxicity testing."
Toxicological Sciences 148.1 (2015): 121-136.
Wilkinson, Grant R., and David G. Shand. "A physiological
approach to hepatic drug clearance." Clinical Pharmacology
& Therapeutics 18.4 (1975): 377-390.
-------