Pharmacophore interactions analysis and prediction of inhibitory activity of 1,7-diazacarbazoles as checkpoint kinase 1 inhibitors: application of molecular docking, 3D-QSAR and RBF neural network

Abstract

In the present study, we mainly focused on new synthesized 1,7-diazacarbazole derivatives (44 active molecules) as Chk1 inhibitors to build 3D-QSAR model. Comparative molecular field analysis (CoMFA) model with three principal components was developed. The relative contributions in building of CoMFA model were 64.41 % for steric field and 35.59 % for electrostatic field. R 2 values for training and test sets of CoMFA model were 0.8724 and 0.7818, respectively, and squared correlation coefficient for leave-one-out cross-validation test (q 2) was 0.6753. To improve the predictive power, a new 3D-QSAR model was developed by using radial basis function network (RBFN) and score of CoMFA interactions energy values as input variables. Scores 1, 2 and 3 were used as input variables, and a RBFN model with seven centers and spread value equal to 95 was developed to create a nonlinear 3D-QSAR model. R 2 values for training and test sets were 0.9613 and 0.8564, and q 2 for leave-one-out cross-validation test was 0.9258. Docking of all molecules to 3DX ligand binding site of Chk1 receptor indicated six interactions as pharmacological interactions between compounds and binding site of receptors. These pharmacological interactions were hydrogen bonding with LEU-15 and GLU-85 in main chain and four van der Waals interactions with LEU-15, VAL-23, TYR-86 and LEU-137 in side chain. CoMFA contour plots were used to design new inhibitors, and inhibitory activity of each compound was predicted by using CoMFA and RBFN models.