QSAR analysis of soil sorption coefficients for polar organic chemicals: substituted anilines and phenols.

Based on descriptors of n-octanol/water partition coefficients (logKow), molecular connectivity indices, and quantum chemical parameters, several QSAR models were built to estimate the soil sorption coefficients (logKoc) of substituted anilines and phenols. Results showed that descriptor logKow plus molecular quantum chemical parameters gave poor regression models. Further study was performed to improve the QSAR model by using artificial neural networks (ANNs). It showed that ANN model with suitable network architecture could make a better agreement between predicted and measured values of the soil sorption coefficients. The quality of the QSAR models confirmed the suitability of ANN to predict the soil sorption coefficients for polar organic chemicals of substituted anilines and phenols.