Prediction of rate constants for the reactions of alkanes with the hydroxyl radicals

The free radicals obtained through an H-atom abstraction pathway (RH + (OH)-O-aEuro cent -> H2O + R-aEuro cent), were used to calculate quantum chemical descriptors for quantitative structure-activity relationship (QSAR) models of rate constants (k (OH)) for reactions of 161 alkanes with hydroxyl radicals ((OH)-O-aEuro cent) in the troposphere. Three quantum chemical descriptors used as the inputs of the support vector machine (SVM) model were selected from 14 quantum chemical descriptors with the genetic algorithm (GA) method together with the multiple linear regressions (MLR) technique. All the descriptors were calculated with the density functional theory (DFT), at the UB3LYP level of theory with 6-31 G(d) basis set. The best prediction results were obtained with the Gaussian radical basis kernel (C = 1, epsilon = 10(-4) and gamma = 0.5). The mean root-mean-square (rms) error for the prediction of k (OH) is 0.314 log units. Our research results indicate that the QSAR model based on GA-MLR and SVM techniques and DFT calculations was accurate and reliable.