Abstract
Using the GUSAR 2013 program, the quantitative structure-antioxidant activity relationship has been studied for 74 phenols, aminophenols, aromatic amines and uracils having lgk(7) = 0.01-6.65 (where k(7) is the rate constant for the reaction of antioxidants with peroxyl radicals generated upon oxidation). Based on the atomic descriptors (Quantitative Neighborhood of Atoms (QNA) and Multilevel Neighborhoods of Atoms (MNA)) and molecular (topological length, topological volume and lipophilicity) descriptors, we have developed 9 statistically significant QSAR consensus models that demonstrate high accuracy in predicting the lgk(7) values for the compounds of training sets and appropriately predict lgk(7) for the test samples. Moderate predictive power of these models is demonstrated using metrics of two categories: (1) based on the determination coefficients R(2) (R(2)(TSi), R(2)(0), Q(2)((F1)), Q(2)((F2)), RmTSi2¯) and based on the concordance correlation coefficient (CCC)); or (2) based on the prediction lgk(7) errors (root mean square error (RMSEP), mean absolute error (MAE) and standard deviation (S.D.)) The RBF-SCR method has been used for selecting the descriptors. Our theoretical prognosis of the lgk(7) for 8-PPDA, a known antioxidant, based on the consensus models well agrees with the experimental value measure in the present work. Thus, the algorithms for calculating the descriptors implemented in the GUSAR 2013 program allow simulating kinetic parameters of the reactions underling the liquid-phase oxidation of hydrocarbons.