Multivariate QSAR Study of Indole β- Diketo Acid, Diketo Acid and Carboxamide Derivatives as Potent Anti-HIV Agents

In this study, a set of novel synthesized - diketo acid, diketo acid and carboxamide derivativesas HIV-1 integrase (HIV-1 IN) was subjected to multivariate QSAR study. Two different variable selection approaches, namely, genetic function approximation (GFA) and multiple linear regression (MLR) used to build the regression models were compared to predict the HIV-1IN inhibition activity. Based on prediction, the best validation model for 5-variable 3’ processing inhibition activity with squared correlation coefficient (R²)= 0.9477, cross validated correlation coefficient (Q²)= 0.9202 and external prediction ability pred_R²= 0.8654. Thisshows that nlowhighest atom weighted BCUTS (BCUTw-1h), minimum E-State for (Strong) Hydrogen bond donors (minHBd), Maximum E-State descriptors of strength for potential Hydrogen Bonds of path length (maxHBint3), Fraction of sp3 carbons to sp2 carbons (HybRatio)and Non-directional WHIM, weighted by atomic masses (WD.mass) were the positive contributors, whereas for 6-variables 3’ processing inhibition activity, parameters R²= 0.9588, Q²= 0.9212 and pred_R²= 0.7364 showed VPC-4, VPC-5, maxHBd, maxwHBa, maxHBint9 andWD.mass contributed positively to the activity. The binding mode pattern of the compounds to the binding site of integraseenzyme was confirmed by two novel parametersr²m(test) and R²p. Y-randomization methods confirmed the model robustness. The results of the present study is useful for designing more potent HIV-1IN inhibitors.