Dr. Oleg Borbulevych will be presenting a talk concerning the Company’s recent work in QM-based X-ray Refinement methods at the American Chemical Society’s National Meeting, Division of Computers in Chemistry, in Denver, Colorado.

Title: A new approach for macromolecular crystallographic refinement: Incorporation of the linear scaling, semiempirical quantum chemical program DivCon into the PHENIX refinement package

Abstract: X-ray crystallography is the primarily technique used to reveal the three-dimensional structure of protein complexes that play a critical role in Structure Based Drug Design. Because of the low ratio of observed data to refined parameters, macromolecular crystallographic refinement at moderate and low resolution relies heavily on the set of known amino acid geometric parameters that ensures the correct stereochemistry of the model. Available programs for macromolecular refinement such as REFMAC, PHENIX or SHELX use simple harmonic oscillator functions to introduce stereochemistry restraints and commonly do not account for electrostatic interactions in the system. This approach inevitably masks important structural details that are often crucial to the understanding of ligand binding within the active site of the protein. To overcome these limitations and achieve a much more realistic description of the protein-ligand geometry, we replaced these stereochemistry restraints with the energy functional derived from quantum-mechanical (QM) treatment. This treatment has been demonstrated with the successful integration of the commercial DivCon ToolKit developed by QuantumBio with the popular Python-based crystallographic package PHENIX. DivCon employs semiempirical QM methods such as AM1, PM3 or PM6 and is based on the divide-and-conquer approach to evaluate the density matrix allowing linear-scaling of the QM problem. As a result, DivCon dramatically decreases the computation costs traditionally associated with QM-based methods, making the application of quantum chemistry for large protein systems feasible.

We proposed a novel protocol to incorporate QM gradients and energy targets into individual (“XYZ”) coordinate refinement step in PHENIX without altering the other refinement stages such as the bulk solvent correction or temperature factor refinement. Furthermore, a user has a choice to use DivCon methods either for the whole structure or a selected region – a ligand and protein active site residues, for example. Based on five test protein structures downloaded from Protein Data Bank (PDB) we report the detailed comparison of the conventional and QM driven refinements. Our preliminary results indicate that incorporation of the QM function not only improves the local geometry but also reduce the R/Rfree factors. For example, the re-refinement of a 17-residue short protein at 2 Å resolution (PDB ID 1S9Z) using the PHENIX/DivCon package indicates a number of significant structural improvements as compared to the conventional PHENIX refinement. Notably, PHENIX/DivCon compared to PHENIX alone improved the peptide bond geometry for the non-standard N terminus residue of that protein and the DivCon driven refinement accurately represents the stereochemistry in this region. Furthermore, the QM approach results in more reasonable H-bond network throughout the protein molecule.