Computational Models for Predicting Human Drug Toxicity

DESCRIPTION (provided by applicant): Advances in genomics, combinatorial chemistry, and high-throughput screening have created a bottleneck at the lead optimization stage of the drug discovery process. Lead optimization includes the redesign of compounds to impart appropriate toxicity properties. Even after lead optimization, greater than 90% of compounds fail in development, and a significant fraction of these failures are due to inadequate toxicity characteristics. Predictive, computational models for toxicity that are based on structure alone allow for a more rapid redesign process and can result in higher quality drug candidates with optimal toxicity properties. We propose to apply two types of approaches-one mechanistic in nature, the other statistical-in a complementary fashion to address the problem of computational toxicity prediction. The mechanistic approach addresses toxicity due to the formation of reactive, metabolic intermediates. It makes use of Camitro' s sophisticated computational models for predicting drug metabolism. The second approach explores the wealth of data contained in the Adverse Event database maintained by the FDA and aims to uncover statistical associations between structures and human toxicity endpoints. Lastly, we intend to combine the two models and to implement our predictive toxicity model with high-throughput screening capacity for access by drug development scientists. PROPOSED COMMERCIAL APPLICATION: Computer-based, high-throughput models for the prediction of human toxicity of new drug candidates are urgently needed by the jpharmaceutical/biotechnology industry in view of the enormous increase in leads identified by new discovery methodologies such as genomics, combinatorial chemistry, and high-throughput compound screening as well as the persisting high cost of drug development. Computational models will enable drug discovery scientists to virtually screen compounds prior to their synthesis to rapidly determine their toxicity profiles and to design drug candidates that have optimal, as opposed to merely adequate, toxicity profiles. Compounds thus selected will have a greater chance of development success and likely lead to better medicines with fewer safety concerns.