Validation and updating of QSAR models for partitioning coefficients of ionic liquids in octanol-water and development of a new LFER model

Since estimating the octanol-water partitioning coefficients (log P) of numerous ionic liquids (ILs) is tedious, time & material consuming and labor intensive, predicting by quantitative structure-activity relationship (QSAR) approach is necessary. Although several researchers presented the QSAR models for the property, validation assessment of the models were not sufficiently performed due to lack of log P dataset. In this study, the log P values of external ILs were measured by a shaking-flask method or collected from literatures. The newly obtained external log P values were applied for the validation study of previous models. In results, it was found that previous models showed rather low predictabilities and/or non-ignorable prediction limits to some IL structures whose anions were not involved in the previous studies. Accordingly, to achieve better predictability, the parameters used for previous modeling were re-selected and also their coefficients were re-calculated by multiple linear regression analysis with an inclusion of the external validation set to previous training set. Moreover, for reasonable understanding of chemical meanings in octanol-water partitioning behavior of ILs, we developed a new prediction model with a few number of descriptors, which has a good accuracy of R² = 0.862 and standard error = 0.564 log units.