LSER Models For Predicting Partition Coefficients Of Organic Pollutants Between Microplastics And Water

Organic pollutants in the aquatic environment can be sorbed by microplastics,and this sorption could affect the environmental fates and ecotoxicities of organic pollutants.The equilibrium coefficients(Kd)of organic pollutants partitioning between microplastics and water are significant for evaluating the adsorbability between microplastics and organic pollutants.On the one hand,experimental determination of Kd values is a hard work full of difficulty;On the other hand,it is also limited by time and cost,and impossible to finish the determination of all compounds one by one.Thus,it is necessary to develop in silico models to predict Kd values.In the present study,linear solvation energy relationship(LSER)model was developed for predicting Kd of organic pollutants between microplastics and water.Excess molar refraction(E),dipolarity/polarizability(S),logarithm of gas-hecadecane partition coefficient(L)were developed to improve the performance of models and obtain LSER parameters conveniently.To validate the performance of E,S models,this study developed Kd models again with predicted E and S values.(1)Based on five molecular descriptors in LSER,LSER models were developed for Kd values of organic pollutants partitioning between polypropylene and sea,polyethylene and sea,and polyethylene and fresh water.For the developed three models,the adjusted determination coefficient(R2adj)range from 0.794 to 0.903;the leave-one-out cross-validated correlation coefficient(Q2LOO)and bootstrap method correlation coefficient(Q2BOOT)range from 0.763 to 0.863 and from 0.720 to 0.804,respectively;the external correlation coefficient(R2ext and Q2ext)range from 0.886 to 0.971 and from 0.825 to 0.954,respectively.Thus,the developed models have high goodness of fit,robustness and predictive ability,and are able to predict Kd values of organic chemicals including polychlorinated biphenyls,polycyclic aromatic hydrocarbons,hexachlorocyclohexanes and chlorobenzenes.(2)Dragon descriptors were calculated to develop QSAR models for E,S,L.For the developed three models,R2 adj range from 0.891 to 0.980;Q2LOO range from 0.890 to 0.980 and Q2 BOOT are 0.800 for all three models;R2ext and Q2 ext range from 0.884 to 0.977 and from 0.884 to 0.977,respectively.Therefore,the developed models have high goodness of fit,robustness and predictive ability,and are able to predict the E,S,L values of diverse chemical classes including alcohols,ethers,phenols,anilines,ketones,aldehydes,chlorohydrocarbons,bromohydrocarbons,nitrobenzenes,esters,phthalates,halogenated benzenes,alkylbenzenes,organic acids,polybrominated diphenyl ethers(PBDEs),biphenyls and chlorinated biphenyls,polycyclic aromatic hydrocarbons(PAHs)and halogenated PAHs,benzamides,ureas,quinones,azobenzenes,organophosphorus compounds,sulfonic derivatives,organic sulfides,organofluorine compounds,organic iodides,organosilicones and aromatic heterocycles.(3)Based on predicted E and S values,QSAR models of Kd values were developed again.For the developed two models,R2 adj is 0.948 and 0.892,respectively;Q2LOO is 0.932 and 0.881,respectively,and Q2 BOOT is 0.752 and 0.683,respectively;R2ext is 0.829 and 0.990,respectively,and Q2 ext is 0.790 and 0.986,respectively.Thus,the developed models have high goodness of fit,robustness and predictive ability,and are comparable to those models based on experimental E and S values.