ABSTRACT
In practice, the standard procedures for the determination of octanol-water distribution constants (and other distribution constants) using the shake flask process are time-consuming and tedious [16]. Further, these methods require significant amounts of pure compounds and have a limited dynamic range ( -3 to 3 log units) determined by the smallest amount of compound that can be accurately determined in either phase. Faster and more reliable methods are highly desirable. Chromatographic procedures using centrifugal partition chromatography [17], thin-layer chromatography [18-20], reversed-phase column liquid chromatography [1-3,21-26), and micellar electrokinetic chromatography [ 4,26-30] have been used for this purpose. Chromatographic methods require very little material (which does not have to be pure), are fast compared to traditional methods, and are relatively easy to automate. These methods are, of course, indirect, based on the construction of a correlation equation [Eq. (1)] between a retention property characteristic of the solute and the chromatographic system for a training set of solutes with experi-
Chromatographic Model for Biopartitioning I 163
mental octanol-water distribution constants (log K0 w). Then, further measurements of the chromatographic retention property in the same system can be used to estimate log Kow values for other compounds:
logKow =P +qlogQ (1) where Q is the retention factor (k) or the retention factor resulting from the extrapolation of log k values to zero organic solvent for binary aqueous mobile phases containing different volume fractions of organic solvent (kw). For thin-layer chromatography, RM or RMW values are used in place oflog k and log kw. In practice, it is important that the structures of the training set and samples are well matched, otherwise incorrect estimates of the distribution constant are obtained. This arises because the system properties that influence log k or log kw are not quite the same as those expressed by the octanol-water distribution constant. When structurally unrelated compounds are correlated via Eq. (1), the agreement obtained is often poor. Recently, a high-throughput, gradient elution method was described to define a chromatographic hydrophobicity index suitable for estimating log Kow for combinatorial libraries [31-33]. This index is based on an earlier solute descriptor ¢0 , defined as the volume fraction of acetonitrile required to achieve an equal distribution of solute between the mobile and stationary phases in reversedphase liquid chromatography [34]. Because a higher volume fraction of acetonitrile is required to achieve the equal distribution (k = 1) for more "hydrophobic" compounds, it is implied that ¢0 is a suitable index of compound hydrophobicity. Correlations between ¢0 or the chromatographic hydrophobicity index and log K0 w are only modest for structurally varied compounds, indicating that the faith in these scales is, in large part, misplaced. From a mechanistic perspective, both ¢0 and the chromatographic hydrophobicity index portray complex processes that tend to mask the system properties involved.
