Prediction of the vapor pressure of some halogenated methyl-phenyl ether (anisole) compounds using linear and nonlinear QSPR methods

Abstract

In this work, several chemometric methods were applied for the modeling and prediction of the vapor pressure (−Log PL°) of halogenated methyl-phenyl ether (anisole) compounds. A genetic algorithm method designed for the selection of variables in the multiple linear regression (MLR) model and also the PCA-ranking technique were chosen as feature selection methods for building a least square support vector machine (LS-SVM) model to predict–Log PL°. The multiple linear regression method was used to build a linear relationship between molecular descriptors and the–Log PL° of these compounds. The LS-SVM was then utilized to construct the nonlinear quantitative structure–activity relationship models. The results obtained using LS-SVM were compared with MLR; this revealed that the LS-SVM model was much better than the GA-MLR model. The root-mean-square errors of the training set and the test set for the PC-ranking-LS-SVM model were 0. 2912 and 0.2427, and the correlation coefficients were 0.9259 and 0.9112, respectively. This paper provides a new and effective method for predicting–Log PL° of organic compounds, and also reveals that PC-ranking-LS-SVM can be used as a powerful chemometrics tool for quantitative structure–property relationship (QSPR) studies.

Keywords:

• ab initio
• electronic structure
• quantum chemistry
• computational chemistry