In silico modelling of selected organic substances adsorption from water onto activated carbon

Quantitative structure-property relationships (QSPRs) are used to simulate the adsorption equilibrium of organic micropollutants (MP) on activated carbon (AC). 46 molecular descriptors of different MPs were calculated. The evaluation for their predictive significance was performed using three adsorption equilibrium datasets from own experiments and taken from the literature. Our analysis showed that almost half of the descriptors have high importance in predicting the adsorption. It varied with the variation of the substances within the datasets. Two types of QSPR models were built: a) based on log D and b) based on the linear free energy polyparameter relationship (ppLFER). Despite the poor internal validation of the log D-based model, two datasets predicted reasonably well. However, the water matrix has led to deviations of up to 1-2 log units in some cases. The quality of the ppLFER model was also moderate. The results indicated future research needs.