Prediction of Gas/Particle Partitioning Coefficients of Semi Volatile Organic Compounds via QSPR Methods: PC-ANN and PLS Analysis
O. Deeba,*, P.V. Khadikarb and M. Goodarzic
aFaculty of Pharmacy, Al-Quds University, P.O. Box 20002 Jerusalem, Palestine
bResearch Division, Laxmi Fumigation and Pest Control Pvt. Ltd., 3, Khatipura, Indore-452 007, India
cDepartment of Chemistry, Faculty of Sciences, Islamic Azad University, Arak
Branch, P.O. Box 38135-567 Arak, Markazi, Iran
Linear and non-linear quantitative structure property relationship (QSPR) models for predicting the gas/particle partitioning coefficients of semivolatile organic compounds were developed based on partial least squares (PLS) and artificial neural network (ANN) to identify a set of structurally based numerical descriptors. Multilinear regression (MLR) was used to build the linear QSPR models using combination of the compounds structural descriptors and topological indices related to environmental conditions such as temperature, pressure and particle size. The prediction results for PLS and ANN models give very good coefficient of determination (0.97). In consistent with experimental studies, it was shown that linear and non-linear regression analyses are useful tools to predict the relationship between the calculated descriptors and gas/particle partitioning coefficient.
Keywords: Gas/particle partitioning coefficient, Semivolatile organic compounds, QSPR, PLS, PC-ANN