Development of quantitative structure-property relationship model for predicting the field sampling rate (R_s) of...

Passive sampling technology has been considered as a promising tool to measure the concentration of environmental contaminants. With this technology, sampling rate (R_s) is an important parameter. However, as experimental methods employed to obtain the R_s value of a given compound were time-consuming, laborious, and expensive. A cost-effective method for deriving R_s is urgent. In addition, considering the great dependence of R_s value on water matrix properties, the laboratory measured R_s may not be a good alternative for field R_s. Thus, obtaining the field R_s is very necessary. In this study, a multiparameter quantitative structure-property relationship (QSPR) model was constructed for predicting the field R_s of 91 polar to semi-polar organic compounds. The determination coefficient (R²_Train), leave-one-out cross-validated coefficient (Q²_LOO), bootstrap coefficient (Q²_BOOT), and root mean square error (RMSE_Train) of the training set were 0.772, 0.706, 0.769, and 0.230, respectively, while the external validation coefficient (Q²_EXT) and RMSE_EXT of the validation set were 0.641 and 0.253, respectively. According to the acceptable criteria (Q² > 0.600, R² > 0.700), the model had good robustness, goodness-of-fit, and predictive performances. Therefore, we could use the model to fill the data gap for substances within the applicability domain on their missing R_s value.