基于理论线性溶解能关系预测有机污染物在PDMS与水中的分配系数

Development of TLSER model for prediction partition coefficients between polydimethylsiloxane and water

通过搜集25个多环芳香烃(PAHs)和84个多氯联苯(PCBs)在聚二甲基硅氧烷(PDMS)上平衡分配系数(log K PDMS-W)的实测值,应用理论线性溶解能关系(TLSER),利用逐步多元线性回归(MLR)方法构建了预测分配系数模型.结果表明:模型具有良好的拟合度(决定系数R 2 adj=0.916)、稳健性(交叉验证系数Q^2 LOO=0.910)和预测能力(外部验证系数Q^2 ext=0.862).同时,模型具有较小的RMSE值(0.264),训练集和验证集中化合物的标准残差|δ|≤3,且杠杆值均小于警戒值h^*(0.103).与前人预测模型相比,该模型具有较少的参数和更加丰富的数据集(109种),化合物的实测值跨度大,应用域也更加广泛,且针对PCBs和PAHs两类化合物有较高的预测精度.因此,本模型可适用于预测在应用域范围内其他与这两类化合物结构相似的有机污染物的K PDMS-W值.

Partition coefficients of the chemical polydimethylsiloxane (PDMS) and water ( K PDMS-W ) were important for determining water phase chemical concentrations using passive sampling devices. As a further exploration of the partition behavior of the compounds, a theoretical linear solvation energy relationship (TLSER) model was established by the stepwise multiple regression (MLR) method. The results show that TLSER models has satisfactory goodness of fit, and predictive capacity with adjusted determination coefficients ( R 2 adj ) of 0.916, cross-validation coefficients ( Q^2 LOO ) of 0.910 , external validation coefficient ( Q^2 ext ) of 0.862, and root mean square error (RMSE) of 0.264 . Moreover, apart from one outlier, all the other compounds in the dataset (109) have the standard residual values greater than ±3, and there are no compounds with leverage values ( h ) higher than the warning value h^* (the warning leverage h^*=0.103). Thus, the model provides a good tool for predicting K PDMS-W values within the applicability domains.