Evaluating computational predictions of the relative stabilities of polymorphic pharmaceuticals

The ability of computational methods to describe the relative energies of polymorphic pharmaceuticals is investigated for a diverse array of compounds. The initial molecular geometries were taken from crystal structures, and energy differences between polymorphic pairs were calculated with various geometry optimization methods. Results using molecular mechanics were compared to experimental calorimetric data and periodic density functional theory (DFT) calculations. The best agreement with experimental heats of transition was shown with energies calculated from geometry optimizations using the Compass force field. Calculations that optimized atomic positions with the Compass force field gave correct energy rankings for all 11 polymorphic pairs studied, with an average deviation of 0.61 kcal/mol from experimental results. These findings suggest that computational methods are poised to predict enthalpy differences between polymorphic forms with levels of accuracy that are quite acceptable when proper approaches are employed.