Abstract
Fuzzy adaptive least-squares (FALS), a pattern recognition method of the correlating chemical structure with activity rating, has been developed to generate quantitative structure-activity relationship (QSAR) models for the drug design and screening hazardous chemicals. It is a novel feature of FALS that the degree to which each sample belongs to its activity rating is given by a fuzzy membership function. The method and its application contributed by the author and his coworkers were briefly reviewed. Using FALS, congeneric QSAR analyses of 29 allergenic α-methylene-γ-butyrolactones and 31 calmodulin inhibitors, and non-congeneric QSAR analyses of aquatic toxicity of 394 organic chemicals and human acute toxicity of 504 compounds were performed to construct predictive QSAR models. Furthermore, predictive accuracies of rodent carcinogenicity of 25 organic compounds issued from U.S. National Institute of Environmental Health Sciences were compared using several methods including a QSAR model with FALS. Finally a revised method of calculationg log P (partition coefficient in octanol/water) was proposed for the QSAR studies. The revised method is not only simple and convenient but also reliable compared with the Rekker method and the Hansch-Leo method.
