Genetic algorithm for CoMFA setting optimization: 3D-QSAR study on α-aminosuberic acid derivatives as anti-cancer compounds

Abstract

In most three-dimensional quantitative structure-activity relationship studies, default SYBYL parameters for comparative molecular field analysis (CoMFA) have been used to derive the models. In this work, a genetic algorithm has been employed for the first time to select the best set of parameters. Three-dimensional quantitative structure-activity relationship analysis of a set of 33 analogues of α-aminosuberic acid as a new generation of histone deacetylase inhibitors was performed. Contrary to the ordinary and region focusing CoMFA models, in genetic algorithm optimized model, H-bond was the preferred field type. Genetic algorithm optimized model showed a better predictive ability (r2pred=0.982, q2LOO=0.828, and q2LMO=0.795) compared with ordinary (r2pred=0.937, q2LOO=0.629, and q2LMO=0.537) and region focusing (r2pred=0.954, q2LOO=0.665, and q2LMO=0.564) models derived by CoMFA default set of parameters. © 2013 John Wiley & Sons, Ltd.