ABSTRACT
The molecular orbital semi-empirical method AM1 was employed to calculate a set of molecular properties (variables) of 22 flavonoid compounds (flavones) with anti-HIV-1 activity and nine new compounds were proposed for anti-HIV-1 activity prediction. Pattern recognition techniques, principal component analysis (PCA), hierarchical cluster analysis (HCA), stepwise discriminant analysis (SDA) and K-nearest neighbor (KNN), were employed in order to reduce dimensionality and investigate which subset of variables could be more effective for classifying the flavones according to their degree of anti-HIV-1 activity. The PCA, HCA, SDA and KNN studies showed that the variables log P (partition coefficient), molecular volume (VOL) and electron affinity (EA) are responsible for the separation between anti-HIV-1 active and inactive compounds. The prediction study was done with a new set of nine analog compounds by using the PCA, HCA, SDA and KNN methods and only one of them was predicted as active against HIV-1.
