Indian Journal of Science and Technology

Abstract

The quantitative structure–retention relationships (QSRR) method is employed to predict the retention times (RTs) of pesticides by molecular descriptors which were calculated by Dragon software. After the calculation molecular descriptors for all molecules, a suitable set of molecular descriptors were selected by using genetic algorithm (GA) and then the data set was randomly divided into training and prediction set. The selected five descriptors were used to build QSRR models with multi-linear regression (MLR) and generalized regression neural network (GRNN) which were built and optimized with intelligent problem solver (IPS) in Statistica 7.1software. Both linear and nonlinear models show good predictive ability, of which GRNN model demonstrated a better performance than that of the MLR model. The root mean square error of cross validation (RMSECV) of the training and the prediction set for the GRNN model was 1.345 and 2.810, and the correlation coefficients (R) were 0.955 and 0.927 respectively, while the square correlation coefficient of the cross validation (Q2 loo) on the GRNN model was 0.951, revealing the reliability of this model. The resulting data indicated that GRNN could be used as a powerful modeling tool for the QSRR studies.
Keywords: Pesticides, Quantitative structure–retention relationship, Genetic algorithm, Multiple linear regression, Retention time, Artificial neural networks.