Comparing approaches for drug-like molecules solubility calculations

Solubility has been recognized as one of the most important properties for designing separation and purification processes of complex molecules, such as active pharmaceutical ingredients. Experimental solubility data are usually needed for performing such design operations. However, frequently data are unavailable due to reduced amounts of sample, time limitations, or inherent complexities with experimental measurements. In such cases, thermodynamic models can be the more theoretically sound tools to generate solubility estimates. In this work, the group-contribution method UNIFAC, and the NRTL-SAC activity coefficient model, are used to correlate and predict solubility in pure and mixed solvents of a set of representative drug-like molecules such as benzoic, salicylic and acetylsalicylic acids, ibuprofen, hydroquinone, estriol, estradiol and resveratrol. Generally, UNIFAC and NRTL-SAC models are able to represent the data, with NRTL-SAC being better for pure solvent solubilities. Solubility dependence with temperature and solvent composition were also taken into account. Whenever possible, the reference solvent approach was also applied, and the results were generally improved with any of the models. The average percent absolute deviations obtained for the representation of solubility data in pure solvents are very satisfactory, but for mixed solvents higher deviations are possible to find.

VIVO