Toward Physics-Based Solubility Computation for Pharmaceuticals to Rival Informatics

We demonstrate that physics-based calculations of intrinsic aqueous solubility can rival cheminformatics-based machine learning predictions. A proof-of-concept was developed for a approach via sublimation thermodynamic cycle, building upon previous work relied several approximations, notably the 2RT approximation, and limited conformational sampling. Here, we apply improvements to our free-energy model with use crystal phonon mode capture contributions vibrational modes crystal. Including these lattice energies computed using model-potential-based ?mol method leads accurate estimates free energy. Combining hydration obtained from either molecular dynamics perturbation simulations or density functional theory calculations, solubilities comparable both experiment informatics predictions are obtained. The application coronene, succinic acid, pharmaceutical desloratadine shows how methods must be adapted adoption different conformations in phases. has flexibility extend applications cannot covered by methods.