Quantitative structure–property relationship of distribution coefficients of organic compounds

The <i>n</i>-octanol/buffer solution distribution coefficient (or <i>n</i>-octanol/water partition coefficient) is of critical importance for measuring lipophilicity of drug candidates. After 4885 molecular descriptor generation, 15 molecular descriptors were selected to develop quantitative structure–property relationship (QSPR) models for distribution coefficients at pH 7.4 (log <i>D</i>_7.4) of a large data set consisting of 1043 organic compounds, which was divided into a training set (600 compounds) and a test set (443 compounds). Support vector machine (SVM) based on genetic algorithm was used to develop a model for log <i>D</i>_7.4 that has coefficient of determination <i>r</i>² of 0.919 for the training set and 0.893 for the test set. The results suggest that the SVM model is accurate in predicting log <i>D</i>_7.4.