Comparison of parametric methods with quantum mechanics based models to estimate Clog P values

The logarithm of octanol-water partition coefficient (log P) is estimated by means of six parametric methods (iLOGP, XLOGP3, WLOGP, MLOGP, SILICOS-IT, OBLogP) and six quantum mechanics (QM) based models. QM models comprise DFT functional choice (B3LYP, M06-X, PBE0, including the xTB parametrization variant), implicit solvent model choice (IEFPCM, SMD, or ALPB), and basis set quality in DFT calculations (6-31G* and 6-311++G**). Several statistics are evaluated to compare Clog P estimates to experimentally measured dataset of 36 molecules picked from Meylan and Howard J. Pharm. Sci. 84 (1995) 83—92. ChemmineR cluster analysis has been performed to confirm structural diversity of molecules in the employed molecular dataset. XLOGP3 performance is the best among all the methods used, while SMD based DFT models showed a clearly competitive results with the remaining parametric methods irrespective of DFT functional or basis set quality.