QSAR study of VEGFR-2 inhibitors by using genetic algorithm-multiple linear regressions (GA-MLR) and genetic algorithm-support vector machine (GA-SVM): a comparative approach

Abstract

The quantitative structure–activity relationship (QSAR) of the novel 4-aminopyrimidine-5-carbaldehyde oxime derivatives as effective and selective inhibitors of potent VEGFR-2 was studied. A suitable set of the molecular descriptors was calculated, and the most impressive descriptors were subsequently selected using genetic algorithm (GA) variable selection approach. To construct robust models, the GA was combined with multiple linear regression and support vector machine, respectively, as GA-MLR and GA-SVM. The predictive quality of the QSAR model was examined for an external set of six compounds, randomly chosen out of 32 compounds in the original data set. The accuracy of the proposed models was further confirmed using cross-validation, validation through an external test set and Y-randomization approaches. Based on the selected descriptors, we have identified some key features in the 4-aminopyrimidine-5-carbaldehyde oxime derivatives that are responsible for potent VEGFR-2 inhibitory activity. The analyses may be used to design more potent 4-aminopyrimidine-5-carbaldehyde oxime derivatives and predict their activities prior to synthesis.