Application of COSMO-SAC to Solid Solubility in Pure and Mixed Solvent Mixtures for Organic Pharmacological Compounds
In this work, we present two open literature databases, the VT-2005 Sigma Profile Database and the VT-2006 Solute Sigma Profile Database, that contain sigma profiles for 1,645 unique compounds. A sigma profile is a molecular-specific distribution of the surface-charge density, which enables the application of solvation-thermodynamic models to predict vapor-liquid and solid-liquid equilibria, and other properties. The VT-2005 Sigma Profile Database generally focuses on solvents and small molecules, while the VT-2006 Solute Sigma Profile Database primarily consists of larger, pharmaceutical-related solutes. We design both of these databases for use with the conductor-like screening model−segment activity coefficient (COSMO-SAC), a liquid-phase activity-coefficient model. The databases contain the necessary information to perform binary and multicomponent VLE and SLE predictions. We offer detailed tutorials and procedures for use with our programs so the reader may also use their own research on our research group website (www.design.che.vt.edu). We validate the VT-2005 Sigma Profile Database by pure component vapor pressure predictions and validate the VT-2006 Solute Sigma Profile Database by solid solubility predictions in pure solvents compared with literature data from multiple sources. Using both databases, we also explore the application of COSMO-SAC to solubility predictions in mixed solvents. This work also studies the effects of conformational isomerism on VLE and SLE property prediction. Finally, we compare COSMO-SAC solubility predictions to solubility predictions by the Non-Random Two-Liquid, Segment Activity Coefficient (NRTL-SAC) model. We find UNIFAC is a more accurate method for predicting VLE behavior than the COSMO-SAC model for many of the systems studied, and that COSMO-SAC predicts solute mole fraction in pure solvents with an average root-mean-squared error (log10(xsol)) of 0.74, excluding outliers, which is greater than the RMS error value of 0.43 using the NRTL-SAC model.