Prediction of soil sorption coefficient of a diverse set of organic chemicals from molecular structure

A correlation study based on simple structural descriptors for predicting the soil sorption coefficient, log K(oc), of a diverse set of 568 organic compounds is presented. Using a training set of 403 compounds, in which the log K(oc) values were in the range 0-6.5, multiple linear regression (MLR) was utilized to build the models. The models were validated using a test set of 165 chemicals not included in the training set. The statistics for a linear regression model with calculated aqueous solubility, log S, were r(2) = 0.80 and s = 0.51 in the training set, and r(2) = 0.76 and s = 0.61 in the test set. The model parameters used allow rapid and accurate calculation of log K(oc) values for a diverse set of organic chemicals, and propose the importance of molecular solubility, lipophilicity, size, flexibility, and ionization for a chemicals' sorption to organic soil material.