UPPSALA, Sweden—A significant challenge in developing drugs is finding compounds that are soluble at therapeutic concentrations. A variety of computational methods have been developed to identify soluble compounds, but researchers at the University of Uppsala worry that these methods relied on training sets particularly sparse of low-solubility drugs and drug-like compounds. Thus, they characterized 15 poorly soluble drugs to determine what factors influenced solubility.
As they report in the Journal of Medicinal Chemistry, the researchers used differential scanning calorimetry and the shake-flask method to analyze compounds ranging in solubility from about 3 nM to 1 µM. They noted immediately that solubility strongly correlated with ClogP values but that the solubility of this group of compounds was about 10-fold lower than would be calculated from a training set of highly soluble drugs.
They then performed linear regression analysis to examine the impact of other properties identified as important for solubility, such as molecular size and. They found that neither melting temperature nor hydrogen bonding parameters—indicators of crystal stability—correlated with solubility. Molecular weight, polarizability, and nonpolar surface area, however, did correlate, suggesting that solvation state rather than solid state factors heavily influence compound solubility.