Computional Framework for Intelligent Predictive Toxicology

The ability to rapidly assess the human health risks posed by a multitude of newly developed chemical compounds is of critical importance. Standard method for predicting toxicity to humans involves in vivo testing in laboratory animals. High cost and timerequirements of the animal studies created an interest in the development of predictive toxicology, in which the toxicity inferences are obtained from a limited number of animal studies with the use of additional, more easily obtainable information such asresults of in vitro studies, and/or physical and chemical properties of the molecules. Although a substantial progress has been achieved in development of methods of predictive toxicology, there is no definitive method (computational or, biological) thatcan be relied upon in the prediction of human toxicity.IAC proposes to develop the iPTE (Intelligent Predictive Toxicology System), an open-architecture computational framework for development and integration of predictive methods. In Phase I, the system architecture will be designed and a prototype will bedemonstrated for a single toxicity end-point. The Phase I research will include implementation of novel neural network classification techniques and development of reliable quantitative measures of the prediction accuracy.The computational system for toxicity prediction is a significant product that will be used for both civilian and military applications. The main function of the system will be to provide rapid screening of newly developed chemicals with respect to humantoxicity. Such screening is required by regulatory requirements for various industries. Additionally, the system may be marketed as an integration and development platform for emerging predictive approaches