Ant colony optimization as a feature selection method in QSAR modeling of anti-HIV-1 activities of 3-(3,5-Dimethylbenzyl)uracil derivatives using MLR and SVM regression

Abstract

A quantitative structure-activity relationship (QSAR) modeling was carried out for the anti-HIV-1 activities of 3-(3,5-dimethylbenzyl)uracil derivatives. The ant colony optimization (ACO) strategy was used as a feature selection (descriptor selection) and model development method. Modeling of the relationship between selected molecular descriptors and pEC50 data was achieved by linear (multiple linear regression-MLR, and partial least squares regression-PLS) and nonlinear (support-vector machine regression; SVMR) methods. The QSAR models were validated by cross-validation, as well as through the prediction of activities of an external set of compounds. Both linear and nonlinear methods were found to be better than a PLS-based method using forward stepwise (FS) selection, resulting in accurate predictions, especially for the SVM regression. The squared correlation coefficients of experimental versus predicted activities for the test set obtained by MLR, PLS and SVMR models using ACO feature selection were 0.942, 0.945 and 0.991, respectively.