Chemometrics Laboratory, Faculty of Chemistry, University of Mazandaran, Babolsar, Iran
In this work, quantitative structure-property relationship (QSPR) approaches were used to predict the redox potential of 42 phenolic antioxidants. The structures of all compounds optimized by the AM1 semi-empirical method and then a large number of molecular descriptors were calculated for each compound in the data set. Subsequently, stepwise multilinear regression was applied to select the most significant and relevant descriptors. The selected descriptors are; the highest occupied molecular orbital energy, the number of hydroxyl groups and harmonic oscillator model of aromaticity. These descriptors were used to develop the multiple linear regression (MLR) and artificial neural network (ANN) models. The values of root mean square error for ANN model were; 0.049, 0.075 and 0.043 for training, internal and external tests sets, respectively, while these values were; 0.061, 0.088 and 0.073, respectively for MLR model. Comparison between these values and other statistical parameters for these two models revealed the credibility of ANN in prediction of redox potential of phenolic antioxidants by using QSPR approaches.