Enhanced Accuracy QSAR-Based Multi-Organ Drug Toxicity Prediction Tool

DESCRIPTION (provided by applicant): The goal of this Phase I project is to demonstrate the feasibility of advancing prototype software (successfully tested on hepatotoxicity) to (1) achieve accurate toxicological endpoint predictions for multiple human or gans, and (2) automate processes thereby reducing user skill requirements. The proposed computer software will offer a faster and more reliable capability for predicting the toxicity of potential pharmaceutical products without animal testing. This tool is expected to be of substantial value to companies developing new pharmaceutical products as it will enable them to screen out drug formulations with unintended negative toxicological consequences prior to spending years and hundreds of millions of dollars on their development. The company's proprietary software uses an advanced quantitative structure-activity relationships (QSAR) approach enhanced through the use of computational neural networks (CNNs), wavelets, modern statistical methods, and molecular me chanics. Jackknife and bootstrap resampling methods are used to avoid bias. The tool can be trained on data obtained via various methods, including microarray, mass spectroscopy, and emerging techniques. The proposed research will allow the company to brin g to market within 3 years a toxicology prediction tool suitable for use by researchers with limited computational chemistry, toxicology, and computer skills. Preliminary prototype testing has demonstrated 94%-98% accurate hepatotoxicology endpoint predict ions. Accuracy increases with increased database size, i.e., client companies with proprietary databases will achieve higher accuracy. The company believes that comparable results can be obtained for multiple human organs and that its evaluation process, w hich currently requires substantial computational chemistry and toxicology expertise by the user, can be automated to enable accurate predictions without those skills. One challenge is eliminating the requirement for user expertise without degrading predic tive accuracy. Phase I research will demonstrate the feasibility of integrating a number of predictive algorithms previously developed by the company with molecular mechanics force field coding to enable predictions to be made by entering basic chemical co mpound formula information. It will demonstrate accurate multiple human organ toxicity predictions (hepatotoxicology and renal toxicology). During the Phase II project, the company will complete predictive and graphical user interface software and user gui des, and commence beta testing by researchers in target customer companies. PUBLIC HEALTH RELEVANCE: This project develops a bioinformatics tool that will enable researchers to better understand the potential toxicities of compounds before advancing them t o clinical trials and exposing humans to them. The proposed tool will help reduce pharmaceutical development costs and the time to bring new drugs to market.