Using Molecular Conformers in COSMO-RS to Predict Drug Solubility in Mixed Solvents

Abstract

This work explores the impact of solute conformers on the conductor-like screening model for real solvents (COSMO-RS) solubility predictions of vanillin and ethyl vanillin in water, short alcohols, and their mixed solvents. Two major conformers of these solutes and changes with solvent polarity were experimentally established by Raman spectroscopy and further confirmed by density functional theory calculations. The COSMO-RS predictions using the individual conformers show a poor description of the solubilities. Estimation with the COSMO-RS default conformer distribution gave better predictions and an intermediate behavior between the predictions obtained using each individual conformer. To further improve the description of the solubilities, the weight of each conformer was fitted to the experimental solid−liquid equilibrium data of the solute in a pure solvent at different temperatures. Better solubility predictions in ternary systems describing solubility maxima were found, suggesting a semipredictive approach to COSMO-RS. This method can predict the liquid−liquid oiling-out effect in the studied binary and ternary systems.