QSAR Models For Predicting Partition Coefficients Of Organic Pollutants Between Passive Sampling Materials And Water

The partition coefficients for organic pollutants between passive sampling materials and water (Kpw) are significant for designing passive sampling devices and calculating water concentrations from the samplers. However, it is difficult to measure Kpw for all potential pollutants since experimental determination of Kpw is generally laborious, time-consuming and expensive. Therefore, it is necessary to develop in silico models for predicting Kpw values. In the present study, multiple linear regression (MLR) analysis was employed to develop quantitative structure-activity relationships (QSAR) models for Kpw of seven passive sampling materials, i.e., polyethylene, polyacrylate and five different silicone rubbers.(1) Based on six molecular descriptors in theoretical linear solvation energy relationships (TLSER), QSAR models were developed for predicting Kpw values of seven passive sampling materials. For the developed models, the adjusted correlation coefficient squares (R2 adj) range from 0.772 to 0.978; the leave-one-out cross-validated Q2 (Q2LOO) and bootstrap method Q2BOOT range from 0.749 to 0.976 and from 0.784 to 0.801, respectively; the external explained R2 ext and Q2 ext range from 0.627 to 0.981 and from 0.612 to 0.977, respectively. The established models, with good goodness-of-fit, robustness and predictive ability, are capable of predicting the Kw values of compounds with various functional groups including>C=C<,-OH,-O-,>C=O,-C=O(O),-C6H5,-NO2,-NH2,-NH- and -X(F, Cl, Br, I). The results indicate that the distribution of organic pollutants between passive sampling materials and water are affected by cavitation and hydrogen bonding.(2) In order to improve the performance of models, Dragon descriptors were calculated to develop QSAR models for KPW. For the developed models,R2 adj range from 0.806 to 0.989, Q2 LOO and Q2 BOOT range from 0.786 to 0.988 and from 0.773 to 0.801, respectively, R2ext and Q2 ext ranged from 0.769 to 0.989 and from 0.757 to 0.982, respectively. The established models, with high goodness-of-fit, robustness and predictive ability, are better than those based on TLSER, and capable of predicting the Kpw values of diverse chemical species including alkanes, alkenes, aromatics, alcohols, ketones, esters and ethers. The dominant molecular structural factors on logKpw of organic contaminants include McGowan volume (Vx), the number of chlorine atoms (nCl), the ring perimeter (Rperim), the number of multiple bonds (nBM), the topological polar surface area with N and O polar contributions (TPSA(NO)), the number of-N(=)= (NddsN), and the number of hydroxyl groups (nROH).