Office of Research and DevelopmentNational Center for Computational Toxicology

David DixDeputy Director, National Center for Computational Toxicology

Application of ToxCast High Throughput Screening to Green Chemical Design

NAS Green ChemistyWashington, DCSeptember 20-21, 2011

This work was reviewed by EPA and approved for presentation but does not necessarily reflect official Agency policy.

Office of Research and DevelopmentNational Center for Computational Toxicology

Too Many Chemicals Too Little Data (%)

Why ToxCast?Why ToxCast?

1

0

10

20

30

40

50

60

Acute Cancer GentoxDev Tox Repro Tox

1

10

100

1000

10000

IRIS TRI PesticidesInerts CCL 1 & 2 HPVMPV

Office of Research and DevelopmentNational Center for Computational Toxicology2

High Throughput Screening 101 (HTS)

96-, 384-, 1536 Well Plates

Target Biology (e.g., Estrogen Receptor)

Robots

Pathway

Chemical Exposure

Cell Population

HTS: High Throughput Screening2

Office of Research and DevelopmentNational Center for Computational Toxicology

ToxCast / Tox21 HTS Chemical Space

3

…

ToxCast

EDSP21

Tox21

1,000 2,000 10,000

500 ToxCast assays

ToxCast EATSassays

Tox21 Assays

Pesticides (active and inert), industrial chemicals, consumer products,marketed and failed pharmaceuticals, food additives, water contaminants, natural human metabolites

Office of Research and DevelopmentNational Center for Computational Toxicology

4

The Home of TFomics TM

attageneattageneattageneattagene

Over 500 ToxCast HTS Assays fromContractors and Collaborators

Compound Focus, Inc.a subsidiary of

EPA/ORD/NHEERL

ToxRefDB ToxCast Human Disease

A. B. C.ToxicityEndpoints

ToxCastAssays

Che

mic

als

Che

mic

als

Office of Research and DevelopmentNational Center for Computational Toxicology

Deepwater Horizon

Oil Exploration Platform Explodes April 20, 2010 • Estimated 4.9 million barrels of South Louisiana Crude released

1.8 million gallons of dispersant used• 1072K surface; 771K subsea• Corexit 9500A (9527 early in spill)

EPA Administrator call for less toxic alternative• Verification of toxicity information on NCP Product Schedule• ORD involvement in assessments of dispersant toxicity

Office of Research and DevelopmentNational Center for Computational Toxicology

Goals of the NCCT Oil Dispersants Project

• Test 8 candidate dispersants for endocrine (ER, AR, TR) activity– Driven by fact that some dispersants contain nonylphenol ethoxylates,

known ER agonists

• Evaluate relative cytotoxicity

• Look for other types of bioactivity using broad in vitro screen

• Return analysis in ~6 weeks

7

Office of Research and DevelopmentNational Center for Computational Toxicology

The Dispersants

8

Sample NameVolume Received Comments Date Received

Manufacturer/Source

Corexit 9500 1 L hazy yellow 11-May-10 Nalco

JD 2000 10 ml clear yellow 27-May-10Ethox Chemicals, LLC

DISPERSIT SPC 1000 10 ml clear amber 27-May-10 PolychemSea Brat #4 10 ml hazy yellow 27-May-10 Alabaster Corp

Nokomis 3-AA 10 mlclear light

color 27-May-10MAR-LEN Supply inc.

Nokomis 3-F4 10 mlclear light

color 27-May-10MAR-LEN Supply inc.

ZI-400 25 ml clear yellow 29-May-10 ZI Chemical

SAF-RON GOLD 500 ml silver iridescent 4-June-10

Sustainable Environmental Technologies, Inc.

Office of Research and DevelopmentNational Center for Computational Toxicology

Dispersant Cytotoxicity Results

9

More potent Less potent

Significantly more cytotoxic (statistically but not biologically)

Bottom line: no significant difference across products

Attagene: HepG2NCGC: Bla ER/ARNHEERL: T47D, MDA, CV1

Office of Research and DevelopmentNational Center for Computational Toxicology

Concentration-Response Profiles for ER

10

Estradiol andNonylphenol compounds

ERa.T=Attagene ERa TRANSERE.C=Attagene ERa CIS

CIS Efficacy less than half TRANS efficacy for reference compounds

Dispersant Positives

TRANS assay efficacy neardetection threshold for these dispersants

Office of Research and DevelopmentNational Center for Computational Toxicology

Biological Pathway Results

11

Xenobioticmetabolism

EstrogenReceptorActivity

Bottom line: 2 products show weak estrogen activity“Not Significant Biologically”

More potent Less potent

Multi-assay pile-upIndication of generalized cell stressPrelude to cytotoxicity

Office of Research and DevelopmentNational Center for Computational Toxicology

Dispersant Conclusions

• Weak evidence of ER activity in 2 dispersants– Seen in single, perhaps over-sensitive assay (1 of 6)– Not of biological significance– Consistent with presence of NPE– Activity only at concentrations >> seen in Gulf after dilution

• No AR activity• No ER activity seen in Corexit 9500• Corexit is in the middle of the pack for cytotoxicity• No worrisome activity seen in other NR assays

12

Analysis of Eight Oil Spill Dispersants Using In Vitro Tests for Endocrine and Other Biological Activity, June 30, 2010 (PDF 47pp) http://www.epa.gov/bpspill/reports/EPADispersantInVitroReport30june2010FINALx.pdfEnviron Sci Technol. 2010 Aug 1;44(15):5979-85.Analysis of eight oil spill dispersants using rapid, in vitro tests for endocrine and other biological activity.Judson RS, Martin MT, Reif DM, Houck KA, Knudsen TB, Rotroff DM, Xia M, Sakamuru S, Huang R, Shinn P, Austin CP, Kavlock RJ, Dix DJ.

Office of Research and DevelopmentNational Center for Computational Toxicology

ToxPi: Prioritizing Chemicals for EDSP21 and Potential Endocrine Activity

Prioritization Index = ToxPi = f(HTS assays + Chemical properties + Pathways)

Bisphenol A Tebuthiuron

Ingenuity pathways

ER

AR

TROther

XME/ADME

KEGG pathways

LogP_TPSA

Predicted CaCO-2

Disease classes

Other NR

Reif et al (2010) Endocrine Profiling and Prioritization of Environmental Chemicals Using ToxCast Data. Environ Health Perspect. Epub PMID: 20826373

13

Office of Research and DevelopmentNational Center for Computational Toxicology

ToxPi Ranking of Chemicals

ToxScore

Reif et al, 2010

14

HPTE

Methoxychlor

Pyrimethanil

Tebuthiuron

309

Che

mic

als

sorte

d by

Tox

Pi

highestlowest

Bisphenol A

Linuron

Office of Research and DevelopmentNational Center for Computational Toxicology

Bioactivity Signature Workflow

ToxCast HTS data (AC50s)

ToxRefDB

UNIVARIATE ASSOCIATIONS• t-test (continuous assay data)• classifier (2x2 contingency table)• Fisher’s exact test (dichotomous data)

MULTIVARIATE MODELS• machine learning• 5-fold cross-validation (80/20 split)• sensitivity analysis (ROC curves)

MULTICELLULAR MODELS• knowledge-base (VT-KB)• computer simulations (VT-SE)• biological inferences / mechanisms

(+)ve (-)ve

(+)ve TP FP

(-)ve FN TN

Office of Research and DevelopmentNational Center for Computational Toxicology

ReproTox Predictions

16

Biol Reprod. 2011 Aug;85(2):327-39. Epub 2011 May 12.Predictive model of rat reproductive toxicity from ToxCast high throughput screening.Martin MT, Knudsen TB, Reif DM, Houck KA, Judson RS, Kavlock RJ, Dix DJ.

Office of Research and DevelopmentNational Center for Computational Toxicology

17

Office of Research and DevelopmentNational Center for Computational Toxicology

Toxicol Sci. 2011 Aug 26. [Epub ahead of print]Predictive models of prenatal developmental toxicity from ToxCast high-throughput screening data.Sipes NS, Martin MT, Reif DM, Kleinstreuer NC, Judson RS, Singh AV, Chandler KJ, Dix DJ, Kavlock RJ, Knudsen TB.

Developmental Toxicity Predictions

Office of Research and DevelopmentNational Center for Computational Toxicology

Systems Modeling: Blood Vessel Development

v Endothelial proliferation,migration & sprouting (VEGF, Chemokines)

• Extracellular matrix degradation(uPAR, PAI-1, MMPs)

v Neovascular stabilization(Ang/Tie2)

19

Environ Health Perspect. 2011 Jul 25. [Epub ahead of print]Environmental Impact on Vascular Development Predicted by High Throughput Screening.Kleinstreuer NC, Judson RS, Reif DM, Sipes NS, Singh AV, Chandler KJ, Dewoskin R, Dix DJ, Kavlock RJ, Knudsen TB.

Office of Research and DevelopmentNational Center for Computational Toxicology

Cell-agent-based models

20

• stochastic cellular behaviors• specified cellular activities• PDE solvers for biochemical gradients• toolbox of morphogenetic processes• executes collective cell behavior• enables emergent properties

CompuCell3D.org

Office of Research and DevelopmentNational Center for Computational Toxicology

21

Molecular Signals and Cellular Behaviors of Angiogenesis

Office of Research and DevelopmentNational Center for Computational Toxicology

22

Angiogenetic field at time=0ECs

ECt

MCIC

SOURCE: Kleinstreuer et al. (in preparation)

Office of Research and DevelopmentNational Center for Computational Toxicology

23

CCL2

CXCL10

Ang1/Tie2

VEGF121

sFlt1

VEGF165

MMPs

Virtual angiogenesis

Office of Research and DevelopmentNational Center for Computational Toxicology

5HPP-33 5HPP-33

24

In vitroSOURCE: Noguchi et al. 2005, Bioorg Med ChemLett. 15 :5509-13.(experimental, 100 uM)

Thalidomide 5HPP-33control

In silicoSOURCE: Kleinstreuer et al. 2011, in preparation

(ToxCastDB, 40 uM)

quantitative predictionemergent from ToxCast HTS data input (AC50s)

TEST CASE: Thalidomide and 5HPP-33

Office of Research and DevelopmentNational Center for Computational Toxicology

25

PFOS

PFOSPFOS

PFOS

PFOA

PFAA_Alt1

PFAA_Alt2PFAA_Alt3PFAA_Alt4

Office of Research and DevelopmentNational Center for Computational Toxicology

26

HUVEC Proliferation: PFAA_Alt2HUVEC Proliferation: PFAA_Alt1

HUVEC Proliferation: PFOS HUVEC Proliferation: PFOA

Inhibition of Endothelial Cell Proliferation

Office of Research and DevelopmentNational Center for Computational Toxicology

PFOA, PFOS, and PFAA Alternatives:in silico vasculogenesis assay

27

PFOA PFOS

PFAA_Alt1 PFAA_Alt2

Office of Research and DevelopmentNational Center for Computational Toxicology

Quantitative Effects:in silico chemical exposure

28

Total Cell Number Total VEGF Concentration

Office of Research and DevelopmentNational Center for Computational Toxicology

29

Green Chemistry and CompTox

• Less expensive, higher throughput approach to hazard prediction

• Prioritization of green alternative chemicals for further development and testing

• High throughput exposure predictions also needed

in vitro testing

ToxicityPathways

Prioritization, Predictiveand Systems Models

Office of Research and DevelopmentNational Center for Computational Toxicology

Virtual EmbryoTom Knudsen Nicole KleinstreuerNisha SipesAmar Singh (LHM)Michael Rountree (SSC)Richard Spencer (EMVL)Rob DeWoskinSid Hunter Stephanie PadillaKelly Chandler

US EPA CompTox

EPA Funded STAR GranteesUniversity of HoustonUniversity of TexasIndiana UniversityUniversity of North Carolina

Virtual LiverImran Shah John WambaughJohn JackWoody Setzer

The Hamner InstitutesRusty ThomasBarbara Wetmore

ToxCastKeith HouckAnn RichardBob KavlockDavid DixMatt MartinDavid ReifRichard JudsonSteve LittleAmy WangPatra VolarathSumit Gangwal