Solubility prediction in non-aqueous binary solvents using a combination of Jouyban-Acree and Abraham models

The Abraham solvent equation coefficients and solute's descriptors were used to develop quantitative structure property relationships (QSPR) for calculating the interaction parameters of the Jouyban-Acree model. The applicability of the proposed methods for predicting solubility of five different solutes dissolved in 194 various non-aqueous binary solvents has been evaluated by mean percentage deviation (MPD). The most accurate method employed experimental solubilities in mono-solvents and water-to-solvent parameters and the overall MPD was 4.7 (±4.0). MPDs of ab inito prediction methods using water-to-solvent and gas-to-solvent parameters were 33.4 (±20.7) and 23.8 (±21.6)%, respectively, and the most accurate predictions were obtained using average predicted solubility approach which produced MPD of 20.4 (±16.8)%. Therefore, using two experimental data the expected prediction error was ∼5 and using the ab initio method was ∼24%.