Lipophilicity Studies on Pyrrolyl‐Acetic Acid Derivatives. Experimental Versus Predicted logP Values in Relationship with Aldose Reductase Inhibitory Activity

Abstract

AbstractThe lipophilicity of a series of pyrrolyl‐acetic acid derivatives, inhibitors of the aldose reductase enzyme, was assessed by direct partitioning experiments in the n‐octanol–water system. Distribution coefficients concerning both the neutral and the anionic species were compared to calculated values generated by commonly used software. The reliability of the predictions was evaluated considering the absolute differences, as well as the correlation between experimental and predicted logP or logD values. The inferior performance of the different algorithms for logD7.4 prediction should be attributed to additional sources of errors and in particular to the nonsystematic interference of ion pairing in partitioning. Although the traditional fragmental system of Rekker as modified in PrologP (logPCDR) was quite successful in logP estimation of the pyrrolyl‐acetic acid derivatives, the nonlinear artificial neural network module ANN2005, implemented in the same software was found to be more consistent in reflecting the features involved in partitioning process. Plot of experimental logP versus aldose reductase inhibitory activity values revealed a definite trend in biological activity and adequate quantitative relationships were further established if, besides logP, Hammett electronic constant σ or Abraham's hydrogen bond acceptor term was included in the regression equation. Among the calculated logP, logPANN05 and ABlogP showed an analogous trend with biological activity, although leading to equations of inferior quality, whereas the other calculated logP values failed to show any regular pattern.