A Practical Prediction of LogPo/w through Semiempirical Electronic Structure Calculations with Dielectric Continuum Model

We examined various combinations of semiempirical and polarizable continuum methods and found that SMD/PM7 provides the highest accuracy in calculating LogP(o/w) The correlation between the experimental and calculated LogP(o/w) is approximately R-2 = 0.74. However, this result is not sufficiently accurate for achieving a practical prediction. To clarify the details of the calculation results, a linear fitting was conducted for each functional group and each atom. The results showed a large variation, indicating a large error. We then conducted a multiple regression analysis using the number of atoms in the molecule, the number of multiple bonds, the ring structure, and the indicator for aromatic nature as explanatory variables and the experimental value of LogP(o/w) as the objective variable; finally, we incorporated a correction to the calculated value of LogP(o/w) . We showed that the accuracy was significantly improved to approximately R-2 = 0.95 when adopting the difference in solvation energy, along with four additional explanatory variables.

Automated summary

By examining various combinations of calculation methods and variables, we found that the SMD/PM7 method provides high accuracy in calculating LogP(o/w). To enhance accuracy, a multiple regression analysis incorporating solvation energy and additional variables significantly improved the results.