In Silico Prediction of Inhibition Activity of Pyrazine–Pyridine Biheteroaryls as VEGFR-2 Inhibitors Based on Least Squares Support Vector Machines

Abstract

A predictive nonlinear model for the inhibition activities for a set of pyrazine–pyridine biheteroaryls, inhibitors of Vascular Endothelial Growth Factor Receptor-2 (VEGFR-2) was developed based on Least Squares Support Vector Machines (LS-SVMs) using molecular descriptors calculated from the molecular structure alone as inputs. Each compound was described by the structural descriptors that encode constitutional, topological, geometrical, electrostatic, and quantum-chemical features. Five relevant descriptors selected by heuristic method were used to build linear and nonlinear Quantitative Structure–Activity Relationship (QSAR) models using Multiple Linear Regression (MLR) and LS-SVMs. Better results were obtained by the nonlinear LS-SVMs model which gave the correlation coefficients of 0.921 and the MSE of 0.046 for the training set. The corresponding correlation coefficient and MSE for the test set are 0.877 and 0.041, respectively. The good performance of LS-SVMs proved this method to be a reliable and promising tool in QSAR analysis and computer aided molecular design. The models developed can be used for further screening of potential VEGFR-2 inhibitors.