Prediction of sediment-water partition coefficients for pesticides in real river based on energy lattice points

Predicting the sorption coefficients of pesticides in the real environment is a complex and challenging task. This complexity arises from the fact that multiple factors, including the characteristics of the contaminants, the composition of the sediment and the prevailing environmental conditions, can exert a significant influence on the sediment-water partition coefficient. Here, 54 commonly-used pesticides were detected in a river. Subsequently, the performance of the extended connectivity fingerprint model and the energy lattice points model in predicting sediment-water partition coefficients was compared. The results indicated that the concentration levels of pesticides in the research region followed a specific order: herbicide concentration (126.35 ± 65.68 ng/L in water and 93.02 ± 39.21 ng/g in sediment) > fungicide concentration (51.88 ± 19.91 ng/L in water and 55.26 ± 18.21 ng/g in sediment) > insecticide concentration (9.25 ± 5.24 ng/L in water and 8.18 ± 4.85 ng/g in sediment) > plant growth regulator concentration (5.68 ± 1.05 ng/L in water and 5.90 ± 2.70 ng/g in sediment). The quantitative structure-property relationship model demonstrated that, in comparison to the traditional extended connectivity fingerprint model, the accuracy of the energy lattice points model in prediction was significantly improved (r² = 0.85 vs. 0.60). This finding suggests that the energy lattice points, when used as a descriptor, can more effectively reflect the non-binding interaction characteristics between small-molecule contaminants and sediments compared to traditional model. And the energy lattice point model can be well applied to the prediction of pollutant adsorption factors in real environments.