Sorption coefficients (Koc) are useful in the prediction of whether a pesticide will remain dissolved in solution or will become adsorbed onto soil particles after its application. Measuring this process experimentally is difficult, expensive and time-consuming. Hence, much effort has been directed toward estimating Koc through statistical modelling. In this study, we investigated the physicochemical properties of pesticides employed by a local sugarcane company, in the northern coastal plain of Paraíba state in Brazil, by using several molecular descriptors, among them, GRid INdependent Descriptors (GRIND). Quantitative assessment of the structure–property relationship (QSPR) model indicated that size, shape, octanol–water coefficient, solubility and the balance between hydrophilic and lipophilic regions, are all relevant to Koc values.