Prediction of AHAS inhibition by sulfonylurea herbicides using genetic algorithm and artificial neural network

Abstract

Acetohydroxyacid synthase (AHAS; EC 2.2.1.6) catalyzes the first common step in branched-chain amino acid biosynthesis. The enzyme is inhibited by several chemical classes of compounds and this inhibition is the basis of action of the sulfonylurea herbicides. The negative logarithm inhibition constant (pKi) of 68 sulfonylurea analogs as inhibitors of pure AHAS using quantitative structure–activity relationship (QSAR) has been calculated. Suitable set of molecular descriptors are calculated and the important descriptors are selected by genetic algorithm and stepwise multiple regression methods. These variables serve as inputs to generated neural networks. After optimization and training of the networks, they are used for the calculation of pKi for the prediction set. Comparison between the results obtained, show the superiority of genetic algorithm over stepwise multiple regression method in feature-selection. For network that used the genetic algorithm for feature-selection methods there are very good agreements between calculated and experimental pKi for data set. The correlation coefficient between calculated and experimental values of pKi for training and prediction set are 0.988 and 0.954, respectively.