We propose to develop a general method for predicting plausible crystalline polymorphs of pharmaceutical molecules. The screening strategy significantly reduces a key problem with existing methods in which global optimization of the total crystalline energy hypersurface does not capture hydrogen bonding well. Our method integrates new theories for identifying polymorphs with screening strategies that ensures all polymorphs contain weak but well-characterized and important intermolecular interactions. Preliminary results of the screening strategy showed that it is very effective in correctly predicting crystal structures of semi-flexible hydrogen-bonded organic molecules: all predicted structures had good hydrogen-bond geometries and optimum packing coefficients-a distinct improvement over the current best known method. We propose to tailor, optimize, and apply this new method towards predicting plausible polymorphs of pharmaceutical molecules. We will incorporate new theories for optimization, and broaden the scope of the screening strategy to extend the method towards more complex problems in crystal prediction (such as solvates, co-crystals, and kinetics). Routine prediction of crystalline polymorphs is our long-term goal. We expect to develop a more expedient and less expensive computational method than is currently achievable for polymorph prediction that is consistent with the way chemists approach solid-state problems.
The importance of predicting polymorphs of pharmaceuticals and other biologically relevant organic molecules is enormous. Routinely predicting plausible polymorphs will broadly impact the cost of design, manufacturing, and characterization processes for these molecules. The proposed method would also benefit research laboratories where the primary focus is on fundamental studies of intermolecular interactions, developing strategies for controlling structure-property relationships, and evaluating new theories of predicting molecular packing arrangements. The methodology will be made commercially available to industry as software.