2D QSAR model using multiple linear regression (MLR)-Genetic algorithm (GA) method for the predication of pKa of some imidazole derivatives

Abstract

pKa is essential in drug discovery as it influences pharmacokinetic and pharmacodynamic properties such as solubility, permeability, absorption and bioavailability of drug molecules, thus crucial in designing stable and effective drug formulations. In this work, 2D molecular descriptors from some 48 selected of imidazole derivatives were used to build robust QSAR models based on multiple linear regressions (MLR) integrated with genetic algorithm (GA) for accurate prediction of pKa. Ten 10 models were developed, and were statistically validated. The R2 values ranged from 0.9131 - 0.9426; adj.R2 (0.9042-0.9367); VIF (11.5078-17.4216); R2-Q2LOO (0.0185-0.0240); SEE (0.1608-0.1874) and MAE (0.1182-0.1472). The model 2 presented with best R2 (0.9426), VIF (17.4217), adj.R2 (0.9367), SEE (0.1475), SDEPLOO (0.1608), Rm2LO (0.8898) and MAE (0.1182) is statistically adjudged the best model with the least deviation from the experimental values; this best model was used to design and predict the pKa values for twelve (12) imidazole derivatives.