Structural alerts to genotoxicity: the interaction of human and artificial intelligence

Abstract

The assignments of 'structural alerts' by Ashby and associates as predictors of genotoxic carcinogenicity were used as the entries for CASE, an artificial intelligence-based structure-activity relational method. CASE, using this human intelligence-based information, was able to derive structural determinants that duplicate the informational content of the structural alerts (sensitivity, 0.974; specificity, 0.948). The CASE-predicted alerts performed as well as the results of the Salmonella mutagenicity assay or the direct application of structural alerts in predicting carcinogens.