Computing intrinsic aqueous solubility of crystalline organic molecules

Accurate computational methods to predict the solubility of crystalline organic molecules in water are highly sought after in many fields of the biomolecular sciences. For example, predictions of solubility are used in the agrochemical and pharmaceutical industries to assess the environmental fate of potential pollutants and the bioavailability of de novo designed drugs, respectively. While a number of statistical Quantitative Structure-Property Relationship (QSPR) solubility models have been built by many groups, including ourselves, development of a method based on chemical theory and molecular simulation has proved challenging. Successful construction of a fully theoretical method would bring both greater insight and systematic improvability, with physically meaningful component effects becoming quantifiable.
Here we show that reasonably accurate values for the intrinsic aqueous solubility of druglike organic molecules are obtainable from theory. We combine sublimation free energies calculated using the crystal lattice minimisation program DMACRYS with hydration free energies calculated using the 3D Reference Interaction Site Model (3DRISM) of the Integral Equation Theory of Molecular Liquids (IET). Intrinsic solubilities of 25 diverse crystalline druglike molecules are computed by our method, obtaining R=0.85 and RMSE =1.45 log10 S units. This is considerably better than results from implicit continuum solvent models such as the polarizable continuum model (PCM). We can fully computationally describe the thermodynamics of the transfer of a druglike molecule from the crystalline solid phase via the gas phase to dilute aqueous solution. Although the 3D RISM free energy functional that we use contains some parameters, our approach is not parameterized against experimentally measured solubilities. On-going work to improve the accuracy of the calculated sublimation thermodynamic parameters will also be discussed.