PREDICTION OF RATE CONSTANTS FOR NITRATE RADICAL REACTIONS USING A SVM MODEL BASED ON DENSITY FUNCTIONAL THEORY

In the present paper, support vector machines (SVMs) are used to develop a quantitative structure-activity relationships (QSAR) model for the reaction rate constants (-logk(NO3)) of 115 heterogeneous organic compounds, through reaction with nitrate radicals (NO3 center dot) in the troposphere. Two quantum chemical descriptors used as the inputs for the SVM model were calculated with density functional theory (DFT), at the B3LYP level of theory with 6-31G(d) basis set. The best predictions were obtained with the Gaussian radical basis kernel (C = 4, epsilon = 0.15 and gamma = 3). The average root-mean square (RMS) error for the prediction of k(NO3) is 0.502 log units, indicating good robustness and predictive ability. The SVM model, reported here, shows better statistical characteristics compared to existing QSAR models.