A quantitative structure–activity relationship study on HIV-1 integrase inhibitors using genetic algorithm, artificial neural networks and different statistical methods

Abstract

In this work, quantitative structure–activity relationship (QSAR) study has been done on tricyclic phthalimide analogues acting as HIV-1 integrase inhibitors. Forty compounds were used in this study. Genetic algorithm (GA), artificial neural network (ANN) and multiple linear regressions (MLR) were utilized to construct the non-linear and linear QSAR models. It revealed that the GA–ANN model was much better than other models. For this purpose, ab initio geometry optimization performed at B3LYP level with a known basis set 6–31G (d). Hyperchem, ChemOffice and Gaussian 98W softwares were used for geometry optimization of the molecules and calculation of the quantum chemical descriptors. To include some of the correlation energy, the calculation was done with the density functional theory (DFT) with the same basis set and Becke’s three parameter hybrid functional using the LYP correlation functional (B3LYP/6–31G (d)). For the calculations in solution phase, the polarized continuum model (PCM) was used and also included optimizations at gas-phase B3LYP/6–31G (d) level for comparison. In the aqueous phase, the root–mean–square errors of the training set and the test set for GA–ANN model using jack–knife method, were 0.1409, 0.1804, respectively. In the gas phase, the root–mean–square errors of the training set and the test set for GA–ANN model were 0.1408, 0.3103, respectively. Also, the R2 values in the aqueous and the gas phase were obtained as 0.91, 0.82, respectively.

Keywords

• Quantitative structure–activity relationship
• Tricyclic phthalimide
• Genetic algorithm
• Artificial neural network