Statistically validated QSAR study of some antagonists of the human CCR5 receptor using least square support vector machine based on the genetic algorithm and factor analysis

Abstract

Quantitative relationships between molecular structures and CCR5 inhibitory activities of forty two 1-amino-2-phenyl-4-(piperidin-1-yl)-butane derivatives were discovered by chemometrics tools including GA-MLR and FA-MLR as linear models and GA-LS-SVM and FA-LS-SVM as nonlinear models. GA-MLR analysis indicated that the topological (X2A) and geometrical (MAXDN) parameters have the most significant influence on the CCR5 inhibitory activity. FA-MLR model describes the effect of topological, geometrical, and quantum indices on the CCR5 inhibitory activity of the studied molecules. A comparison between the developed statistical methods revealed that FA-LS-SVM represented superior results and it could predict about 95 % of variance in the inhibitory activity data. The resulted models were validated for generalization and predictability by leave-one-out (LOO) cross-validation method. External validation showed high predictability of the calibration models. The predictability of the obtained QSAR models was also investigated by the Tropsha and Roy proposed criteria. Further validation by Y-randomization method confirmed that the obtained models were not due to a chance correlation. The applicability domain of the models was defined by leverage value. None of the studied compounds were outside the domain of the models.