[Computerized logP prediction using fragment methods].

Lipophilicity, expressed by the logarithm of octanol/water partition coefficient (logP) is an important physico-chemical property in rational drug design. Beside the experimental determination, the calculation of logP based on the chemical structure is frequently necessary. This has led to the development of numerous logP prediction methods. In the present paper the fragment type approaches and their computer softwares are surveyed (Table I.). The compilation is extended to the introduction and evaluation of a recently developed method of Meylan and Howard [21]: Atom/Fragment Contribution, AFC method (KOWWIN for Windows, software) which possesses the unique option, the Experimental Value Adjusted, EVA logP prediction. The author compared the highly precise experimental logP values of 28 drugs measured in her laboratory with calculated logP values obtained by four approaches: KOWWIN, CLOGP, PROLOGP, ACD/logP. The best prediction was found as follows in decreasing order: KOWWIN (r = 0.983), CLOGP (r = 0.978), PROLOGP/Combined (r = 0.953), ACD/logP (r = 0.942), PROLOGP/Atomic5 (r = 0.940), PROLOGP/Rekker (r = 0.909). The limits of current logP prediction methods (intramolecular H-bond formation, tautomerization, conformation changes, etc.) and the promising future of the molecular lipophilicity potential, MLP, [42] in drug design is also discussed.