Correlations of log Values with log Poct and Other log P Scales

Although log k or preferably log kw values are good lipophilicity indices per se, very often a direct correlation or even conversion into log Poct values is performed. The mean reason is that one would like to have a comparison with standard experimental lipophilicity values or to calculated log P values, which also refer to the 1-octanol/water system.

Figure 4. Typical log Pact versus log k„ plot. Relationships are found for closely related compounds, namely phenylureas (•), phenoxycarboxylic acids (O) and phenoxycarboxylic acid methyl esters (□) (Reprinted with permission from [7], copyright 1983, Elsevier.)

log Poet amphiprotics and H-donors

H-acceptors non H-bonders

log k

Figure 5. Generalized plot of log Poc[ against log k*: differences in lipophilicity scales reflect hydrogen-bonding capability in both systems.

Despite the fact that log P values are often claimed to correlate well with log k or log K values, this is only so for very closely related compounds (see Fig. 4). And even within an apparently structurally similar series of compounds, subtle effects such as differences in hydrogen-bonding capacity can be found (see Fig. 5). In a series of investigations Yamagami and coworkers formulated the following equations [18, 19]:

In the expressions the three HB terms are indicator variables expressing hydrogen bonding effects of substituents, defined as follows: HBco = 1 for esters and amides, HBa = 1 for H-acceptors other than COOR and CON, and HBAM = 1 for H-donor substituents. This equation illustrates well that the differences between lipophilicity scales can be expressed in terms of hydrogen-bonding capacity (Eq. (19)). Nevertheless, a method based on fragment indicator values is not general enough.

A limitation of HPLC lipophilicity measurements is caused by the inhomogenous nature of the stationary phase resulting in specific interactions with certain solutes. Thus, it is expected that certain column types reflect more the partioning behavior in octanol/water than others. In a comparison of four different columns (see sections 5.5.2) it was found that the ABZ column gives the best correlations with log Poa. However, for more complex structures, such as most biologically active compounds, the correlation becomes less significant (see Table 2). This limited data set is illustrative for what is often seen in practice: log k„ values for small closely related compounds correlate well with log Paa values; for more structurally diverse drugs the correlation is less to nonexistent.

It has recently been found that the differences between log P values measured in two different systems gives relevant information on hydrogen-bonding capacity of the molecules, which is related to their transport properties through membranes (see Chapter 14). It would be of great interest to find similarly a Alog k or Alog parameter derived from two different columns. Generally log values on ODS colums corre-

log k = a log P + hcaHBco + hAHBA + hAMHBAM + pa, + c

asonably well with octanol/water, but poorly with alkane/water partition coefficients. Interestingly, the reverse seems to be true for C18-derivatized PS-DVB (Act-I) columns [6]. However, this observation needs further verification.

0 0

Post a comment