Toxicity prediction of chemicals based on structure-activity relationships

The prediction system of chemical toxicity has been developed by means of structure-activity relationships (SARs) based on the computerized fact database (BL-DB). Number and ratio of elements, side chains, bonding, position and microenvironment of side chains were used as the structural factors of chemicals for toxicity prediction. Such information was obtained from BL-DB by Wiswesser line notation. In the present study, the Salmonella-microsome assay of the mutagenicity was chosen as indicative of the target toxicity of chemicals. A set of chemicals specified with mutagenicity data was retrieved, and the necessary information was extracted and transferred to the working file. Rules of the relation between characteristics of chemical structure and the assay results are extracted as parameters far rules by experts on rearranged data set. There are analysed statistically by discriminant analysis and prediction and the rules were evaluated by the elimination method. For example, eight kinds of rules to predict Salmonella/microsomal assay were constructed, and currently results of the assay on organic compounds can be predicted above 90% accuracy. (C) 1998 Elsevier Science Ireland Ltd. All rights reserved.