Comparative QSAR studies of CYP1A2 inhibitor flavonoids using 2D and 3D descriptors

Comparative Quantitative Structure Activity Relationship (QSAR) analyses have been performed with 21 naturally occurring flavonoids for their inhibitory effects on cytochrome P450 1A2 enzyme using two-dimensional (topological, structural, and thermodynamic) and three-dimensional (spatial) descriptors. The chemometric tools used for the analyses are stepwise multiple linear regression, partial least squares, genetic function approximation, and genetic partial least squares. The data set was divided into a training set (n = 15) and test set (n = 6), based on K-means clustering technique applied on standardized two-dimensional descriptor matrix, and models were developed from the training set compounds. The best model (genetic partial least squares model using two-dimensional descriptors) was selected based on the highest external predictive R(2) (R(2)(pred)) value (0.840) and the lowest root mean square error of prediction value (0.351). The developed QSAR equations suggest the importance of the double bond present at 2 and 3 positions and requirement of absence of hydroxyl substituent or glycosidic linkage at 3 position of the 1,4-benzopyrone nucleus. Furthermore, the phenyl ring present at 2 position of the 1,4-benzopyrone ring should not be substituted with hydroxyl group. Moreover, hydroxyl groups present at 5 and 7 positions of the benzopyran nucleus should not be glycosylated for good cytochrome P450 1A2 enzyme inhibitory activity.