Comparison of the Accuracy of LogP and LogD
Calculations for 22 Drugs

Eduard A. Kolovanov and Alanas A. Petrauskas
 

I. Abstract

It has been shown that the logD parameter is calculated more accurately than the logP parameter. This is explained by the fact that the experimental values of logD are more accurate than logP. LogD is always obtained from direct measurements under the mild conditions, pH being close to 7.4. LogP is either re-calculated from logD, or obtained under the extreme conditions, pH being above 10.0 or below 2.0. These results reveal that both the experimental and calculated logD values are more suitable parameters for characterization of drug hydrophobic interactions than the experimental or calculated logP values.

II. Introduction

In a previous article¹⁾ we evaluated experimental logP data for 22 drugs and compared 6 different logP calculation methods. This and other similar works^2-5) have been published based on assumption that logP is the best physicochemical parameter for describing hydrophobic interactions between drug and receptor binding center or during drug transport to receptor. However all the drugs analyzed in our previous article are the electrolytes ionized under the physiological conditions. LogP is a "theoretical" parameter which characterizes neutral (unionized) molecules only and therefore may not be directly related to the drug transport and interactions with receptor due to solute ionization. For this reason logD is much more suitable parameter for correlating drug biological action, since it takes into account drug ionization. These considerations are also confirmed by the recent comparison of logP and logD correlations with enzyme inhibition and binding constants.⁶⁾ In all cases logP produced correlation coefficients R<0.90, whereas logD produced R>0.99. So the only advantage of logP parameter over logD is that the latter seems to be more difficult to calculate as accurately as logP is.

The purpose of this work was to compare the accuracy of logP and logD calculations for the 22 drugs from our previous article.¹⁾ If the results of these two correlations are comparable, then we can conclude that the logD parameter can be calculated as accurately as logP.