Prediction of polydimethylsiloxane-water partition coefficients based on the pp-LFER and QSAR models

Obtaining accurate measurements of the partition coefficients between sorbent materials and water is of major importance for the analysis of the adsorption behavior and dissolved concentrations of organic compounds in the environment. In the passive-sampling approach, polydimethylsiloxane (PDMS) has a wide range of applications. Therefore, we established a poly-parameter linear-free energy relationship (pp-LFER) and a quantitative structure-activity relationship (QSAR) model to predict the log K_PDMS-w values for a large dataset of 290 organic chemicals from 11 diverse classes. For the pp-LFER model, E (excess molar refractivity), A (molecular H-bond donor ability), V (McGowan volume), and B (the H-bond acceptor properties) were introduced as the main correlated variables. However, the obtained model is much limited in terms of acquiring available descriptors. For this reason, we developed a QSAR model, and CrippenLogP (Crippen octanol-water partition coefficient), RNCG (Relative negative charge-most negative charge/total negative charge), ATSC4e (Centered Broto-Moreau autocorrelation-lag4/weighted by Sanderson electronegativities) and GATS6p (Geary autocorrelation-lag6/weighted by polarizabilities) were selected as the significant parameters. The predictive power and functional reliability of the presented models were confirmed with validation methods as described in previous studies. The adjusted determination coefficients (R²_adj) of 0.851 and 0.922 and leave-one-out cross-validated (Q²_LOO) of 0.841 and 0.907 revealed that the models have good predictive power and generalizability. Thus, the proposed models are simple yet accurate tools for predicting the log K_PDMS-w values and providing new insights to further understand the adsorption mechanism of organic compounds.