Development of quantitative structure-activity relationships and computer-aided drug design

Abstract

Recent development of quantitative structure-activity relationships (QSARs) and computer-aided drug design contributed by the author and his coworkers was briefly reviewed. Fuzzy adaptive least-squares (FALS), a pattern recognition method for analysing structure-activity rating data to generate QSAR models was developed. A novel feature of FALS is that the degree to which each sample belongs to its activity class is given by a fuzzy membership function. Using FALS, non-congeneric QSAR analyses of carcinogenicity, mutagenicity, and six kinds of pharmacokinetic properties of miscellaneous organic chemicals were performed to coinstruct predictive models for drug design. In these QSAR analyses, the values of log P (partition coefficient in octanol/water) calculated by the simple method of Moriguchi et al. were used as the descriptor for hydrophobicity. The method of Moriguchi et al. is not only simple and convenient but also reliable in the application to 22 drugs selected by Rekker et al. compared to the Rekker method and the Hansch-Leo method. Lastly, a heulistic search method for active conformers using molecular mechanical conformational analysis and principal component analysis was proposed.