The development of methods that predict the toxicity expected for any untested chemical of interest, based on knowledge of the chemical's structure, substructure, physicochemical behavior, or biological behavior in short-term or in vitro tests, can lead to a better understanding of mechanisms that underlie toxic modes of action and of factors that influence inter-species extrapolation. They can also greatly enhance management of the selection of chemicals for testing and the identification of tests that most need to be conducted, thereby reducing the cost of testing programs and lessen their dependence on animals. Past efforts to develop predictive methods based on NTP standardized tests for mutagenicity (four in vivo short-term test), Balb/c-3T3 mouse embryo cell cytotoxicity and transformation, and carcinogenicity indicate that systematic analyses of organ-specific, histopathology data from NTP subchronic and chronic toxicity studies is a most promising direction in which to extend this work. The abstraction and compilation of such organ-specific histopathology data from NTP Technical Reports for the 200 most-recently studied chemicals is in progress and will be completed during the next year. The organ-specific, histopathology data, along with other parameters, will be analyzed using new computer-based approaches developed in the field of artificial intelligence that are especially capable of identifying patterns and correlations embedded in large amounts of seemingly unrelated data. This interdisciplinary work will be accomplished through collaborations with Dr. G. Klopman at Case Western Reserve University, using a particularly promising knowledge-based computer system called CASE (computer assisted substructure evaluation) and with Dr. D. Bahler at North Carolina State University, using artificial intelligence approaches. Collaborations will also be developed through the DERT's Small Business Innovative Research (SBIR) program.
