Conventional macromolecular crystallographic refinement relies on stereochemistry restraints and a rudimentary energy functional to ensure the correct geometry of the model of the macromolecule, along with any bound ligand(s), within the experimental, X-ray density. Traditionally, these highly approximate methods lack explicit, rigorous terms for electrostatics, polarization, dispersion, hydrogen bonds, and other interactions, and they often rely on pre-determined parameters to capture the a priori understanding of the structure. In order to address this deficiency and capture a more complete understanding of the structure, we have developed a fully automated approach for macromolecular refinement based on a two layer, QM/MM (ONIOM) scheme implemented within our DivCon Suite which has been "plugged in" to two mainstream crystallographic packages: PHENIX[1] and BUSTER. This implementation consists of one or more "region layer(s)" characterized using linear-scaling, semi-empirical quantum mechanics, coupled with a "system layer" encompassing the rest of the protein described with a molecular mechanics functional[2].
Armed with a more accurate tool, we not only gain a better understanding of overall protein:ligand structure, but we can also use X-ray data to correctly determine active site tautomer/protomer states[3] and water site locations.
In this talk, we will discuss these methods and explore their impact in the context of binding affinity prediction and structure-based drug discovery.