PerspectiVe

Future of ToxicologysPredictive Toxicology: An Expanded View of“Chemical Toxicity”

Ann M. Richard*National Center for Computational Toxicology, Mail Drop D343-03, U.S. EnVironmental Protection Agency,

Research Triangle Park, North Carolina 27711

ReceiVed May 31, 2006

A chemistry approach to predictive toxicology relies on structure-activity relationship (SAR) modelingto predict biological activity from chemical structure. Such approaches have proven capabilities whenapplied to well-defined toxicity end points or regions of chemical space. These approaches are less well-suited, however, to the challenges of global toxicity prediction, i.e., to predicting the potential toxicity ofstructurally diverse chemicals across a wide range of end points of regulatory and pharmaceutical concern.New approaches that have the potential to significantly improve capabilities in predictive toxicology areelaborating the “activity” portion of the SAR paradigm. Recent advances in two areas of endeavor areparticularly promising. Toxicity data informatics relies on standardized data schema, developed forparticular areas of toxicological study, to facilitate data integration and enable relational exploration andmining of data across both historical and new areas of toxicological investigation. Bioassay profilingrefers to large-scale high-throughput screening approaches that use chemicals as probes to broadlycharacterize biological response space, extending the concept of chemical “properties” to the biologicalactivity domain. The effective capture and representation of legacy and new toxicity data into mineableform and the large-scale generation of new bioassay data in relation to chemical toxicity, both employingchemical structure information to inform and integrate diverse biological data, are opening exciting newhorizons in predictive toxicology.

Contents

Introduction 1257Local vs Global 1258Bringing More Information To Bear onToxicity Prediction

1258

Toxicity Data Informatics 1258Bioassay Profiling 1259

Conclusion 1261

Introduction

Predictive toxicology, from a chemistry vantage point, usesa structure-activity relationship (SAR) built on available testdata to predict the potential biological activity of a chemicalbased solely on its molecular structure and computed properties.Predictive toxicology approaches based on SAR modeling areentirely dependent on and limited by the conventions andreference data of toxicology studies, in terms of both thechemicals chosen for study and the experimental end pointsconsidered most informative for toxicological inferences. How-ever, SAR modelers typically have had little prospectiveinfluence in defining either the chemical scope or the measured

* To whom correspondence should be addressed. Tel: 919-541-3934.Fax: 919-685-3263. E-mail: richard.ann@epa.gov.

OCTOBER 2006

VOLUME 19, NUMBER 10

© Copyright 2006 by the American Chemical Society

10.1021/tx060116u CCC: $33.50 © 2006 American Chemical SocietyPublished on Web 09/02/2006

parameters of toxicological studies. The entire enterprise oftoxicology, historically organized around specific categories ofresponses in whole animals, is undergoing profound change.Trends effecting this change are three-fold: (i) large-scale datageneration involving more fundamental, interdisciplinary tech-nologies (driven by genomics and proteomics); (ii) the emer-gence of extensive public information resources to support theseefforts; and (iii) greater investigative focus on chemical andbiological mechanisms that underlie toxicological responses anddisease states (1). Given its intimate reliance on toxicityreference data, the field of predictive toxicology is poised toundergo a correspondingly dramatic shift in focus and approach.The most significant drivers of this change are not the tradetools of chemists and SAR modelers, i.e., new chemicaldescriptors, new data analysis tools, or more sophisticatedalgorithms for defining chemical similarity, or domain ofapplicability and validity of models, although in all of theseareas there have been significant advances in recent years.Rather, the greatest drivers for change will be in the elaborationand enrichment of the “activity” portion of the SAR paradigm.

Local vs Global

Focused computational chemistry investigation has provenits ability to elucidate well-defined biochemical or mechanisticoutcomes for a series of closely related chemical structures. Suchmechanistically based SAR models are most successful whenapplied to well-defined categories of toxicities (2) or to narrowlydefined subsets of chemicals and toxicity measures or targetinteractions (3, 4). Hence, SAR modeling performs best whenapplied locally with respect to either chemistry or biology.

The most pressing challenges confronting predictive toxicol-ogy are more global in nature. In the environmental regulatorydomain is the need to prioritize testing (5), reduce testing (6),and eliminate testing entirely in some cases (7) for a widediversity of chemicals posing a broad range of potentialtoxicological concerns. In the pharmaceutical domain, the goalis to anticipate a spectrum of potential adverse effects of drugcandidates in humans based largely on surrogate in vitro andanimal test results. Some success has been achieved with theso-called “global SAR approaches” to predict broadly acrosschemical space for a particular toxicity end point. These oftenrely on a set of structural alerting features (i.e., chemicalfragments) to grossly represent many regions of chemical spaceassociated with a toxicity end point (8). However, the problemof predicting the potential toxicity of a never-before-seenchemical has largely defied accurate and reliable solutions andis considered by many to be beyond the reach of SARapproaches that rely exclusively on chemical structures andassociated properties. Global models for a toxicity end pointhave proven of value in providing guidance and approximateprediction estimates; for example, if a known structural alertingfeature for that end point is present, the chemical has a greaterthan chance probability of being associated with that toxicity.However, these models frequently fail to predict activitymodulations within local structure-activity space (e.g., quite afew, but not all, aromatic amines are carcinogenic) or to predictoutside the chemical space of the training set, where in bothcases sufficient chemical test examples do not exist to adequatelyinform the model (9).

A major challenge to be met by any successful global SARpredictive toxicity model is the sufficient characterization andcoverage of local domains or “neighborhoods” of similarchemicals. When defined exclusively in chemical terms, thismeans that sufficient toxicity test data within chemical analogue

neighborhoods are needed to enable confident SAR inferencesand sufficient test coverage across chemically diverse neighbor-hoods is needed to enable global SAR predictions. It is implicitlyassumed that this broad chemical coverage, in turn, will spanthe full range of biological mode-of-action neighborhoods inrelation to the ultimate adverse effect measured. Given that theserequirements are with respect to each reference toxicologicaltest end point of interest, these broad data requirements are aprescription either for endless chemical toxicity testing or forfailure. Furthermore, given the diversity and extent of chemicalspace and the lack of full or even partial mechanistic definitionfor most toxicological end points of regulatory or human healthconcern, coverage requirements of all constituent local modelsare unlikely to be met except in the rarest of circumstances.

A more fundamental problem with the concept of a globalSAR model able to confidently predict the potential for toxicityin uncharted areas of chemical space is that chemical structurespace is typically more rugged and discontinuous than corre-sponding biological activity space when the latter is representedby an integrated whole animal response (this may not be thecase, however, when activity is represented at the biochemicalor genomic response level). A basic tenet of SAR investigationthat works well and often enough to be useful is that smallchanges in chemical structure lead to correspondingly smallchanges in biological activity. However, a biological organismis a complex system that frequently imposes discontinuity ontochemical space, such as when small structural changes triggermetabolic activation or a target interaction and genomictranscriptional response, with a correspondingly large impacton the ultimate biological outcome. In addition, the SARparadigm is notoriously bad at projecting into areas of chemicalstructure space not sufficiently covered by the training set ofcompounds. One could argue that unrealistic expectations havebeen placed on SAR approaches to tackle these global toxicityprediction problems; there are simply not enough test data, andlikely never will be, to support this SAR global toxicityprediction objective as a viable way forward.

Bringing More Information To Bear on ToxicityPrediction

If predictive toxicology is to significantly advance in capabil-ity and utility, it must broaden its effective utilization of bothlegacy and new chemical toxicity test data across levels ofbiological organization and traditional toxicity study areas. Inaddition, to break free of the known limitations of chemistry-exclusive SAR toxicity prediction models, a new paradigm isneeded for defining more functionally meaningful chemical/biological activity neighborhoods that are informed by integratedbiological measures and that bear closer relation to the ultimatetoxicity end point being predicted. Major advances that holdgreat promise for effecting these changes are occurring, largelyin the public domain, in the areas of toxicity data informaticsand bioassay profiling.

Toxicity Data Informatics

Toxicity data informatics (Figure 1) refers to: (i) increasingstandardization and digitization of legacy toxicity data andmigration of these data into the public domain; (ii) developmentand population of database standards and models that facilitatedata integration and enable relational exploration and miningof data across both historical and new areas of toxicologicalinvestigation; and (iii) association of these data with chemicalstructuressa universal metric spanning chemical toxicology

1258 Chem. Res. Toxicol., Vol. 19, No. 10, 2006 Richard

domain that enables structure analogue searching and SARhypothesis generation.

A data model, or schema, consists of controlled vocabulary,standardized data descriptions, and a hierarchical field structurethat reflects layers of study data and biological summarizationand organization (10). Data models are being developed withthe involvement of toxicology domain experts in areas such asgenetic toxicity, chronic toxicity, and developmental toxicity(11, 12). A data model enables read-across exploration andmining of toxicity study. Such models are essential first stepsfor more effective data utilization and encourage both theefficient capture of newly generated data and the migration oflegacy data archives to enrich public toxicity data resources[from the historical literature as well as from governmentregulatory agencies, such as the Food and Drug Administrationand the Environmental Protection Agency (EPA)]. With thebuilt-in flexibility to collapse and summarize study data tovarious quantitative and qualitative “end points”, data modelsoffer a route to new testable hypotheses relating structure tobiological activity (10) and potentially rich content input to next-generation prediction models (13).

Populated toxicity data models, by virtue of being top-levelchemically indexed, i.e., searchable by chemical name, CASregistry number, or structure, also can be easily linked to thelarger world of chemically indexed public toxicity information.In contrast, many currently available biological data resourcesderived from chemical exposure testing are not yet chemicallyindexed. Prominent examples include the largest of the publicmicroarray data repositories, GEO and ArrayExpress (14, 15),in which even minimal chemical annotation, i.e., a chemicalname or CAS registry number, is not required in data submis-

sion, and neither chemical name nor CAS number is presentedas a field search option for accessing microarray experimentswithin the databases at the time of this writing. As a result,these experimental results remain effectively isolated in datasilos that serve limited study domain interests (16), and thechemical coverage and inventory within these databases arelargely unknown. Chemical structure indexing is advancing ontothese islands, however, and is proving to be a major andeffective integrator of biological information across the Internet(17). This is perhaps best exemplified by the PubChem Project(18), a fully public data model that provides on-line structuresearchability and full and open data access to millions ofchemically indexed bioassay records from various public anduser-deposited sources. The DSSTox database project is anexample more specifically targeted to toxicology study areas,providing summary representations of activity potentially usefulfor SAR modeling applications, with standardized chemicalstructure annotation enabling read-across searchability over awide range of chemical toxicity information (19). These andother public initiatives, such as the new InChI text structurerepresentation (20), are moving us closer to the reality of a fullystructure-searchable “chemical semantic web” (21). Such ad-vances will enhance the ability to gather or mine a broad rangeof information on a chemical and its analogues from publicInternet resources in support of chemical toxicity predictionefforts.

Bioassay Profiling

Bioassay profiling refers to: (i) broad characterization of achemical in terms of biological response space, i.e., extending

Figure 1. Sample construction of toxicity data model, including DSSTox structure annotation standards, ToxML toxicity data schema with hierarchicalsummary toxicity layers, and importation of these elements into a chemical relational database that supports data mining, modeling, and prediction(screenshot of Leadscope Toxicity Model).

PerspectiVe Chem. Res. Toxicol., Vol. 19, No. 10, 20061259

chemical “properties” to molecular functional and cellular levelsin the biological domain; (ii) bioassay data typically generatedin high-throughput and high-content assays for a wide spectrumof biological targets and cellular responses; and (iii) data thatprovide new information and have the potential, alone or incombination with other data, to serve as bioindicators of anintegrated toxicological response in the whole animal of concern.

A new paradigm for improving toxicity prediction is begin-ning to emerge that resembles the directed high-throughputscreening (HTS) approaches more commonly employed insupport of drug discovery in the pharmaceutical industry (22).In this approach, large numbers of chemicals representingchemical diversity space are employed as probes of biologicalactivity. This is accomplished through the use of a wide varietyof HTS bioassay screens measuring biochemical activity andcell function, effectively generating structure-activity informa-tion de novo across biological function space. Anchoring ofthese chemical-bioactivity profiles to appropriate referencetoxicological data provides the interpretive context for develop-ing predictive models. This new approach to data generationand prediction differs in two very fundamental respects frompast efforts. First, SAR-based predictive toxicology approachestraditionally have exerted little to no influence on whichchemicals or limited areas of chemical space were to undergotesting. In the new HTS approaches, chemicals are chosen tocover large areas of chemical diversity space and to broadlyprobe biological function space without a priori assumptions(23). Second, the nature of the data being generated at thechemical:biological interface represents a projection of chemicalstructure space into the realm of biological functional activity.As a result, there is potential for breaking free of the inherentlimitations of a purely chemical similarity perspective anddefining perhaps fewer or more coherent biological activityneighborhoods in relation to an integrated whole animalresponse.

Covell and co-workers have considered the use of multidi-mensional bioassay profile information for growth inhibitionactivity across a series of 60 tumor cell lines in relation to thelarge NCI chemical structure space (24). In their words, “Theconsiderable chemical and biological diversity inherent in thesedata offers an opportunity to establish a quantifiable connectionbetween chemical structure and biological activity. We find thatthe connection between structure and biological response is notsymmetric, with biological response better at predicting chemicalstructure than vice versa.” Similarly, Fliri and co-workers atPfizer have examined the use of bioassay profiling as meansfor classifying chemicals according to common modes of activityand therapeutic biological response categories (25). Theirfindings provide evidence for coherent biological responsepatterns that overlap to some degree with chemical similarityconsiderations but that are not fully determined or anticipatedby chemical similarity alone. These efforts to date have beenlargely proof-of-principle, focused on particular tumor cell lines,in the case of the NCI, or on limited numbers of commerciallyavailable HTS assays and pharmaceutical type chemicals. Somesuccess has also been reported in deriving genomic signaturesto predict selected toxicity end points of concern to thepharmaceutical industry (26). Although promising, these ge-nomics approaches have limited applicability to global predictionwhen signatures are derived from proprietary microarraydatabases, when testing is limited mostly to pharmaceutical typecompounds, when the chemical structure space of the trainingset and compound of interest is not considered in prediction,and when the application of the signature to prediction requires

use of either the commercial service or the microarray experi-ments with comparable protocols to this service.

A very large public HTS initiative taking shape under theNIH Molecular Libraries & Imaging Roadmap Initiative (MLI)(27) has the potential to greatly enrich data resources and ourbasic understanding of chemical structure in relation to bio-functional activity space. To support the HTS component ofthis effort, the NIH Molecular Library Small Molecule Reposi-tory (28), consisting of upward of 70000 chemicals spanninglarge regions of chemical diversity space, is to undergo screeningin several hundreds, perhaps ultimately thousands, of bioactivityand biofunctional HTS assays (29). This testing program is wellunderway, with HTS data currently being generated across theMolecular Libraries Screening Center Network (MLSCN) (27)for an ever-growing list of HTS assays. As they are produced,these data are being made fully publicly available in PubChem(18), the public data distribution arm of the NIH MLI effort.

Given the tremendous data generation capability enabled bythe new HTS technologies, what is the practical outlook forthese new data impacting our ability to predict or screen forchemical toxicity? Success will depend on many additionalfactors. Are the chemical structures chosen for HTS sufficientlyrepresentative of environmental and industrial chemicals thatcomprise the bulk of reference toxicity data sets, to which theHTS data must be anchored? Are the HTS assays sufficientlyrepresentative of the bioactivity and biofunctional domains mostrelevant to the wide array of potential toxicity end points ofconcern? Are the achievable HTS concentrations appropriateto capture toxicologically relevant effects, i.e., to minimize falsenegatives? Will HTS generate sufficient numbers of actives andunique profiles across tested chemical space to be useful foractivity discrimination? Also, will the chosen HTS assayssufficiently capture requirements for metabolic activation, eitherthrough sufficient sampling of chemical space (to includepossible metabolites) or by explicit inclusion of metaboliccapabilities in the assays?

Despite these varied concerns, the opportunity to enhancepredictive capabilities from HTS data generation is simply toogreat to ignore. As a result, both EPA and the NIEHS NationalToxicology Program are embarking on their own directed HTSefforts, as well as actively engaging with the NIH ChemicalGenomics Center (NCGC is the intramural screening center inthe MLSCN) to expand and improve the value of the HTSresults specifically in relation to chemical toxicity. The NTPhas submitted 1408 chemicals (30, 31), a large portion of thisset overlapping with the on-line NTP reference toxicity database(32), for testing by the NCGC in several toxicity-related HTSscreens. This preliminary set of chemicals and assays has beenrun by the NCGC at expanded dose ranges (up to 15 dosedilutions), and a broader set of chemicals and HTS assays areunder consideration for further testing. Within the EPA NationalCenter for Computational Toxicology’s new ToxCast initiative(33), HTS and bioassay profiling are envisioned to be majorcomponents of toxicity “forecasting”. The EPA ToxCastprogram is collaborating with the NTP and NCGC efforts, aswell as pursuing customized and commercial HTS testing ofseveral hundred pesticidal actives, for which reference toxicitydata across a number of end points of regulatory concern areavailable. Ultimately, the objective is to extract patterns ofbioactivity profiles, possibly in combination with other typesof data, that are predictive of integrated biological responses.Given the large amounts of HTS data to be generated in theseefforts, there is additionally the potential to derive companionSAR models for particular HTS end points, augmenting HTS

1260 Chem. Res. Toxicol., Vol. 19, No. 10, 2006 Richard

data or enabling the prediction of bioactivity profiles in theabsence of HTS data (34).

Conclusion

There are many exciting new developments in the world ofpredictive toxicology that should engage this community andgive rise to significant optimism for the future (see Figure 2).First and foremost, and long overdue, are efforts to make betteruse of the reference toxicity data that we already have, to makeit publicly available, structure-searchable, mineable, moresuitable for modeling, and better integrated across domains oftoxicological study. Second are large public efforts directedtoward the generation of new bioassay profiling data of potentialrelevance to toxicology, where the active engagement oftoxicologists, chemists, and modelers is essential if these dataare to be effectively integrated with existing data and optimallyutilized. Last, but not least, is recognition of the key role to beplayed by chemical structures in integrating diverse biologicaldata, rationalizing such data through enhanced SAR models,and serving as effective probes of biological activity space.

To comment on this or other Future of Toxicology perspec-tives, please visit our Perspectives Open Forum at http://pubs.acs.org/journals/crtoec/openforum.

Acknowledgment. This manuscript was reviewed by theU.S. EPA’s National Center for Computational Toxicology andapproved for publication. Approval does not signify that thecontents necessarily reflect the views and policies of the Agency,nor does mention of trade names or commercial productsconstitute endorsement or recommendation for use.

References

(1) Liebler, D. C. (2006) The future of toxicologysThe poisons within:Application of toxicity mechanisms to fundamental disease processes.Chem. Res. Toxicol. 19(5), 610-613.

(2) Cronin, M. T. D., and Schultz, T. W. (2001) Development ofquantitative structure-activity relationships for the toxicity of aromaticcompounds toTetrahymena pyriformis: Comparative assessment ofthe methodologies.Chem. Res. Toxicol. 14(9), 1284-1295.

(3) Yu, S. J., Keenan, S. M., Tong, W., and Welsh, W. J. (2002) Influenceof the structural diversity of data sets on the statistical quality of three-dimensional quantitative structure-activity relationship (3D-QSAR)models: Predicting the estrogenic activity of xenoestrogens.Chem.Res. Toxicol. 15(10), 1229-1234.

(4) Aptula, A. O., Patlewicz, G., and Roberts, D. W. (2005) Skinsensitization: Reaction mechanistic applicability domains for structure-activity relationships.Chem. Res. Toxicol. 18(9), 1420-1426.

(5) DSL (Domestic Substance List): An initiative of Canadian Environ-mental Protection Act to categorize substances to categorize eithermanufactured in or imported chemicals (23000) for their toxicity orpotential to become toxic by 2006; http://www.ec.gc.ca/substances/ese/eng/dsl/dslprog.cfm.

(6) EPA High Production Volume Challenge Program; http://www.epa.gov/chemrtk/hpvchmlt.htm.

(7) REACH (registration, evaluation, and authorization of chemicals);http://europa.eu.int/comm/environment/chemicals/reach.htm; http://ecb.jrc.it/REACH/; 7th Amendment for cosmetics: A ban on the useof live animals for testing cosmetic products and ingredients by 2013;http://www.eceae.org/english/cosmetics.html.

(8) Kazius, J., McGuire, R., and Bursi, R. (2005) Derivation and validationof toxicophores for mutagenicity prediction.J. Med. Chem. 48, 312-320.

(9) Benigni, R. (2005) Structure-activity relationship studies of chemicalmutagens and carcinogens: Mechanistic investigations and predictionapproaches.Chem. ReV. 105, 1767-1800.

(10) Yang, C., Richard, A. M., and Cross, K. P. (2006) The art of datamining the minefields of toxicity databases to link chemistry to biology.Curr. Comput.-Aided Drug Des. 2(2), 135-150.

(11) ToxML: A publicly available XML standard and controlled vocabularyfor representing toxicity data (2006); http://www.leadscope.com.

(12) Using developmental toxicity data in SAR model development. Projectof the ILSI Research Foundation, Risk-Sciences Institute (2006); http://rsi.ilsi.org/Projects/DevToxSAR.

(13) Yang, C., Benz, R. D., and Cheeseman, M. A. (2006) Landscape ofcurrent toxicity databases and database standards.Curr. Opin. DrugDiscoVery DeV. 9 (1), 124-133.

(14) National Center for Biotechnology Information (NCBI) Gene Expres-sion Omnibus (GEO) (2006); http://www.ncbi.nlm.nih.gov/geo/.

(15) European Bioinformatics Institute (EBI) ArrayExpress (2006); http://www.ebi.ac.uk/arrayexpress/.

(16) Wishart, D. S., Knox, C., Guo, A. C., Shrivastava, S., Stothard, P.,Chang, Z., and Woolsey, J. (2006) DrugBank: A comprehensiveresource forin silico drug discovery and exploration.Nucleic AcidsRes. 34, D668-D672.

(17) Richard, A. M., Gold, L. S., and Nicklaus, M. C. (2006) Chemicalstructure indexing of toxicity data on the internet: Moving towards aflat world. Curr. Opin. Drug DiscoVery DeV. 9 (3), 314-325.

(18) National Center for Biotechnology InformationsPubChem projectstructure-relational database and browser (2006); http://pubchem.ncbi.nlm.nih.gov.

(19) EPA DSSTox, Distributed Structure-Searchable Toxicity DatabaseNetwork website (2006); http://www.epa.gov/ncct/dsstox/.

(20) The IUPAC International Chemical Identifier (InChI): InternationalUnion of Pure and Applied Chemistry, Research Triangle Park, NC(2006); http://www.iupac.org/inchi/.

(21) Coles, S. J., Day, N. E., Murray-Rust, P., Rzepa, H. S., and Zhang,Y. (2005) Enhancement of the chemical semantic web through theuse of InChI identifiers.Org. Biomol. Chem. 3(10), 1832-1834.

(22) Whitebread, S., Hamon, J., Bojanic, D., and Urban, L. (2005) In vitrosafety pharmacology profiling: an essential tool for successful drugdevelopment.Drug DiscoVery Today 10(21), 1421-1433.

(23) Strausberg, R. L., and Schreiber, S. L. (2003) From knowing tocontrolling: A path from genomics to drugs using small molecularprobes.Science 300, 294-295.

(24) Wallqvist, A., Huang, R., Thanki, N., and Covell, D. G. (2006)Evaluating chemical structure similarity as an Indicator of cellulargrowth inhibition.J. Chem. Inf. Model 46, 430-437.

(25) Fliri, A. F., Loging, W. T., Thadelo, P. F., and Volkmann, R. A. (2005)Biological spectra analysis: Linking biological activity profiles tomolecular structure.PNAS 102(2), 261-266.

(26) Fielden, M. R., Eynon, B. P., Natsoulis, G., Jarnagin, K., Banas, D.,and Kolaja, K. L. (2005) A gene expression signature that predictsthe future onset of drug-induced renal tubular toxicity.Toxicol. Pathol.33, 675-683.

(27) NIH Molecular Libraries & Imaging Roadmap Initiative (2005); http://nihroadmap.nih.gov/molecularlibraries/.

(28) NIH Molecular Libraries Small Molecule Repository; http://mlsmr.discoverypartners.com/MLSMR_HomePage/.

(29) Austin, C. P., Brady, L. S., Insel, T. R., and Collins, F. S. (2004)NIH Molecular Libraries Initiative.Science 306, 1138-1139.

(30) National Toxicology Program High-Throughput ScreeningProgram (2006); http://ntp.niehs.nih.gov/ntpweb/index.cfm?objectid)05F80E15-F1F6-975E-77DDEDBDF3B941CD.

(31) Smith, C., Collins, B., Tice, R., Wolf, M. A., and Richard, A. M.(2006) DSSTox National Toxicology Program High ThroughputScreening Structure-Index File: SDF file and documentation, Launchversion NTPHTS_v1a_1408_10Apr2006; http://www.epa.gov/ncct/dsstox/sdf_ntphts.html.

Figure 2. Multidimensional and broadly integrated activities andinformation domains that will support improved predictive toxicologycapabilities.

PerspectiVe Chem. Res. Toxicol., Vol. 19, No. 10, 20061261

(32) NTP On-line Bioassay Database, Study Data searches; http://ntp-server.niehs.nih.gov/index.cfm?objectid)72016020-BDB7-CEBA-F3E5A7965617C1C1.

(33) Dix, D. J., Houck, K. A., Martin, M. T., Richard, A. M., Setzer, R.W., and Kavlock, R. J. (2006) The ToxCast program for prioritizingtoxicity testing of environmental chemicals.Toxicol. Sci.(Forum), inpress.

(34) Poroikov, V. V., Filimonov, D. A., Ihlenfeldt, W. D., Gloriozova, T.A., Lagunin, A. A., Borodina, Y. V., Stepanchikova, A. V., andNicklaus, M. C. (2003) PASS biological activity spectrum predictionsin the Enhanced Open NCI Database Browser.J. Chem. Inf. Comput.Sci. 43, 228-236.

TX060116U

1262 Chem. Res. Toxicol., Vol. 19, No. 10, 2006 Richard