Soil Sorption Estimator (SSE): Read-Across-Based Predictions of Soil Organic Carbon Partition Coefficient (Log K oc) and Validation Using Multiple External Sets

Abstract The persistence and bioaccumulation of organic chemicals are major concerns for environmental and chemical regulatory bodies. Among various properties, the soil organic carbon-water partition coefficient (Koc) is a key parameter that governs the environmental fate of these organic chemicals. Experimental values of soil sorption potential are available for only a limited number of chemicals due to the high cost and labor required for experiments. In the present study, we have developed both quantitative structure–property relationship (QSPR) and quantitative read-across structure–property relationship (q-RASPR) models for the fast and efficient prediction of Log KOC values of new or untested chemicals. The statistical and predictive qualities of our models (R2 > 0.84, QLOO2 > 0.84, QF12 > 0.87, MAEtr < 0.34, MAEtest < 0.44) depict their efficiency in predicting Log KOC values for new organic chemicals. The developed models were validated on 5 external data sets, and their quality was also compared with that of previously developed models. The comparison results depict the superiority of our developed models, showing low prediction errors. A web-based predictive platform, “Soil Sorption Estimator (SSE)” has been developed to compute Log KOC of chemicals with applicability domain (AD) information, providing insight into the reliability of predictions.