A simple QSPR model for predicting soil sorption coefficients of polar and nonpolar organic compounds from molecular...

A quantitative structure-property relationship (QSPR) model is developed to predict the logarithm of the soil sorption coefficient of 82 organic compounds. The data set contains polar and nonpolar, saturated, unsaturated, aliphatic, aromatic, and polycyclic aromatic compounds covering a log K(oc) range from about 1 to 6 log units. The best correlation equation, containing only five constitutional descriptors (number of benzene rings, molecular weight, number of N, O, and S atoms), predicts log K(oc) with a squared correlation coefficient of 0.94, having a standard deviation, s, of 0.33. The model is validated with an external set of 43 compounds not included in the training set. The descriptors involved in the model can be obtained easily from the molecular formula without any further calculation; therefore, the model is ready to use by environmental scientists with no background in quantum chemistry or chemical graph theory or when no software is available.