New Horizons in Toxicity Prediction. Lhasa Limited Symposium Event in Collaboration with the University of Cambridge
Introduction : David Hawkins, Lhasa Limited; Robert Glen, University of Cambridge
Toxicology is a multidisciplinary science that examines the adverse effects of chemicals on organisms. It is a rapidly developing area with many new scientists entering the field. There is a shift from primarily in vivo animal studies to in vitro assays, in vivo assays with lower organisms, and computational modeling for toxicity assessments.1 The conference explored various current approaches to toxicity prediction, covering and comparing the tools and methods available today, uses by regulators, industry and academia, and a look at emerging areas and technologies.
Current Approaches for Toxicity Prediction
Pharmaceutical perspective - Edwin Matthews, FDA
The goal of the Food and Drug Administration (FDA) Center for Drug Evaluation and Research (CDER) ComTox program is to be able to predict accurately chemical toxicities with in silico software for all toxicological and clinical effect endpoints of interest to the U.S. FDA. A benefit would be substantially to reduce, replace and refine the need for animal toxicological testing in establishing the safety of chemical substances. The Informatics and Computational Safety Analysis Staff (ICSAS), part of CDER's Office of Pharmaceutical Science, is facilitating an orderly transition to a new in silico testing paradigm. This is being articulated in parallel to the current Organization for Economic Cooperation and Development (OECD) and European Union QSAR efforts, but there are substantial strategic differences in these approaches, e.g., ICSAS employs commercial software products (which are excluded from the EU effort); freeware is only used for special applications. There is a commitment to global QSAR and expert systems. Commitment to global QSAR means that new models must be added all the time.
Three software platforms are already validated and two additional platforms are being validated. The programs are Derek for Windows and Meteor; Leadscope FDA Model Applier and Predictive Data Miner; MCASE/MC4PC and META; Prous BioEpisteme and Integrity; and QSARIS (formerly MDL-QSAR, now from Scimatics).2 Insilicofirst (founding members Lhasa, Leadscope, MCASE and Molecular Networks)3 is a collaborative endeavor working to develop a computational prediction system to support the environmental safety assessment of chemicals.
The FDA has several reasons for using more than one QSAR software program. None of the programs has all the necessary functionalities, and none has 100% coverage, sensitivity, and specificity. All of the programs are complementary and can be used for consensus prediction strategies. Moreover, FDA cannot endorse a single (Q)SAR program. Components of the multiple platform strategy are predicted or experimental value; bioavailability; structural analogues; coverage (i.e., domain of applicability); metabolites; weight of evidence predictions (combining predictions from multiple QSAR programs); and mechanism of action.
Matthews listed some unmet needs in QSARs and expert systems. These include integrated fragment and descriptor paradigms and 3D descriptors; QSARs based upon pure active ingredient (PAI) and metabolites; QSARs for drug-drug interaction, for animal organ toxicities, and for regulatory dose concentration endpoints (e.g., lowest observed effect level (LOEL) and no observed effect level (NOEL)); and expert system rules for toxicities of substances such as biologicals which cannot be predicted by QSAR. Other unmet needs are databases of pharmaceutical off-target activities, of pharmaceutical Investigational New Drug (IND), of confidential business information and of regulatory dose concentration endpoints; integration of FDA and Environmental Protection Agency (EPA) archival data; and advanced linguistic software to extract data.