Intelligent Toxicology Prediction System

The objective is to determine how two artificial intelligence technologies (intelligent data discovery and neural networks) can be used to provide predictions of a chemical's potential toxicity and carcinogenicity through biological data from short-term tests. The project team expects to find that predictive patterns exist within databases of short-term in vitro tests, subchronic tests, and carcinogenicity test. The team also hopes to find that a neural network can be developed using these relationships and patterns to predict a chemical's potential toxicity and carcinogenicity without the need for specific, long-term organ toxicity studies. The resulting product would benefit both government and the private sector by reducing the time and cost of testing and the dependence on animal testing. The methodology used will begin with identification of available data sources and selection of those to be used in the study. The data will be acquired and prepared for use by data discovery software that finds previously unknown patterns, relationships and correlations in the data contained in the database. These findings will then be used to assist in designing and training of a neural network for toxicological prediction. The neural network will be tested and its predictive performance evaluated and reported.