Abstract

A new improved group contribution model that predicts the n-octanol/water partition coefficient (logP) is described. A combined parameter set that contains 153 basic parameters, 41 extended parameter and 14 molecular surface/property descriptors was generated from a training database of 8320 chemicals. The model achieved significant improvement after modifying the traditional group contribution equation by using a three dimensional steric hindrance modulator. The predictive ability of this model was accessed by calculating the logP values of a test set of 1667 ordinary organic chemicals and a set of 137 drug-like chemicals that were not included in the training database.

Keywords: n-octanol/water partition coefficient, group contribution, klogp, calculation