Shuffling multivariate adaptive regression splines as a predictive method for modeling of novel pyridylmethylthio derivatives as VEGFR2 inhibitors

Abstract

The vascular endothelial growth factor receptor (VEGFR2) is an attractive target for the development of novel anticancer agents. Molecular docking and quantitative structure–activity relationship (QSAR) were used to investigate how inhibitors’ chemical structures relate to the inhibitory activities. The molecular docking studies show that at least one hydrogen bond with LYS866 residue is one of the essential requirements for the optimum binding of a series of 42 pyridylmethylthio inhibitors. The obtained QSAR model indicates that the inhibitory activity can be described by solvent-accessible molecular surface area, topological electronic indices, local dipole index, steric interaction, and hydrogen bonding energies between the receptor and the inhibitors. Furthermore, several validation methods were used to evaluate the predictive capacity of the generated models. The satisfactory results (R ²_L25 %O  = 0.819, Q ²_LOO  = 0.838, R ²_p  = 0.866, RMSE_LOO = 0.315, and RMSE_L25 %O = 0.337) suggest that the models exhibited considerable predictive power which can be used in prediction of activity of new pyridylmethylthio inhibitors. Also the docking analysis showed that the interaction of the inhibitors with residues ALA879, ASP(1044, 1026), LEU880, PHE843, and LYS866 plays an important role in the activities of the inhibitors.