Use of partition coefficients in combination with the molecular formulas of solutes to predict physicochemical properties with improved accuracy

The partition coefficient (logP) between water and an organic solvent has been used widely to predict the physicochemical properties of organic compounds. However, many physicochemical properties cannot be predicted accurately with logP. In this study, a theoretical approach is used to investigate how the molecular sizes, hydrogen (H)-bond acceptors and donors of solutes affect the correlation between their physicochemical properties and logP. We find that the correlations, which are usually strong for nonpolar solutes, are weakened by the H-bond acceptors and/or donors of solutes. When the predictive variable, S m, which can be easily calculated from the molecular formula of solutes, is included in the models for the prediction of physicochemical properties from logP, the statistical reliabilities and predictive power of the models increase remarkably. This strategy to improve the accuracy of the prediction of physicochemical properties using logP does not require additional experimental data or complicated calculations and can be easily used by researchers. It is expected that this strategy can find widespread application in accurately predicting the effects of organic compounds on human and environment.