Analysis of the biological activity of antidiabetic compounds using CoMFA models

As a prerequisite for obtaining the bioactive conformations of the 54 molecules of interest, the docking technique was employed using GOLD1 program. After obtaining the alignment, the PLS (Partial Least Squares) technique was used to perform the CoMFA analyzes and to generate the values of r2 (regression coefficient) and q2 (correlation coefficient with crossvalidation) using the Sybyl X2 package. The internai validations were performed using the leave one out (LOO) method. The parameters tested for the distances between grid points were: 0.5; 1.0; 1.5 and 2.0A; and the weight values (w) were: 0.3, 0.5, 0.7 and 0.9; checking without/with region focus. The parameter that was the most adequate for the construction of the model consisted of distance equals to 1.0 A and the weight of 0.5 between the points of the grid. The next step was to separate the training and test sets to verify the predictive capacity of the model (42 molecules in the training and 12 structures in the test sets ). The MATLAB3 package was used for splitting the compound sets. The best model was the one obtained with w = 0.5 and d = 1.0A (q2 = 0.863 and r2 = 0.991 ); the stereochemical and electrostatic fields were calculated with Lennard-Jones and Coulomb potentials. By analyzing the contour maps for the least (molecule 1) and the most active (molecule 40), the contribution regions of these ligands for the biological activity in the DPP-4 enzyme were identified, which will aid in the design of new bioactive ligands.