[Prediction of -octanol/water partition coefficient of strongly ionized compounds by ion-pair reversed-phase liquid chromatography with silica-based stationary phase].

The -octanol/water partition coefficient (log ) is an important parameter to characterize the overall hydrophobicity of organic compounds. Reversed-phase liquid chromatography (RPLC) has been recommended as an effective method for the indirect determination of log by the Organization for Economic Cooperation and Development (OECD). Using RPLC, most studies focus on the determination of log or the apparent -octanol/water partition coefficient (log ) of neutral compounds and weakly ionized compounds. However, the experimental log or log values of strongly ionized compounds have rarely been reported. In our previous work, the experimental log of strongly ionized compounds could be determined well by ion-pair reversed-phase liquid chromatography (IP-RPLC) on an octadecyl-poly(vinyl alcohol) column using the log -log model established by different types of model compounds. However, the universality of this strategy for different chromatographic columns has yet to be verified. In this study, the retention behavior of neutral compounds, phenolic acids, carboxylic acids, sulfonic acids, and some amphoteric compounds was systematically investigated on a silica-based C18 column (150 mm×4.6 mm, 5 μm) via ion-suppressed RPLC (IS-RPLC) and IP-RPLC, respectively. In the IS-RPLC mode, methanol and 20 mmol/L ammonium dihydrogen phosphate buffer (pH 7.0) were used as the mobile phase to perform isocratic elution at different methanol ratios. The log values of the test compounds were obtained using the linear solvent strength (LSS) model. Neutral compounds, weakly ionized phenolic acids and benzene carboxylic acids were then used as model compounds to establish the log -log model. The quantitative structure-retention relationship (QSRR) model, including structure-related descriptors like the charge () and Abraham solvation parameters ( and ), exhibited much better correlation than the unary linear regression model between log and log . The log (log under pH 7.0) values of 19 ionized compounds were then determined by the model; the determined compounds were used as model compounds and validation compounds in IP-RPLC. In the IP-RPLC mode, besides methanol and ammonium dihydrogen phosphate buffer, the mobile phase also contained tetrabutylammonium bromide, as an ion-pair reagent. The retention behaviors of all tested compounds conformed well with the LSS model even under IP-RPLC, with a log - linear correlation coefficient ( ) greater than 0.99. The log -log model was then established using 62 compounds as a mixed model set, including neutral, weakly ionized, and strongly ionized compounds. Similarly, by introducing , , and , the log -log model showed good linearity, with greater than 0.94. Comparing the log -log model established on the silica-based C18 column in this work with that established on a poly(vinyl alcohol)-based C18 column in our previous work, , , and contributed more to the model in this work, indicating there was a greater secondary effect on the silica-based column. To confirm the reliability of the log -log model, three different types of acidic compounds were used as validation compounds. The predicted log of the three ionic compounds was very similar to that determined by the shake-flask method (SFM)/slow stirring method (SSM) or IS-RPLC method in this work, confirming the reliability of the model. Based on the above results, the log values of eight strongly ionized compounds were predicted by IP-RPLC. The findings suggested that IP-RPLC is a promising method to predict the experimental log of strongly ionized compounds, and that the conventional silica-based C18 column offers more flexible options in log determination.