First steps towards quantum refinement of protein X-ray structures

Publication Type:
Quantum Simulations of Materials and Biological Systems, 2012, pp. 87 - 120
Issue Date:
Filename Description Size
10.1007%2F978-94-007-4948-1_6.pdfPublished version503.88 kB
Adobe PDF
Full metadata record
© Springer Science+Business Media Dordrecht 2012. Using standard force-fields and empirical restraints in protein refinement has proven to be a key tool in X-ray protein structure determination. However, detailed analysis of the resulting structural models sometimes reveals chemically unreasonable features, originating in many cases from the representation of multiple configurations using some averaged structure. Quantum chemical methods and computational capabilities have now come to the point at which full quantum refinement of protein structure is feasible, but only complete (meaning real ensembles of) chemical structures may be considered. Density functional theory (DFT) is currently the most popular quantum chemical approach but a large number of approximate functionals are available and most of these do not correctly describe the biologically important London dispersion effects. For small molecules it has been shown that efficient dispersion corrections can overcome this problem, without additional computational effort. We show that this is also the case using linear-scaling dispersion-corrected DFT to refine protein X-ray structures. The study considers the effect on the R factors (i.e. the agreement between modeled and observed diffraction data) when DFT is used to optimize atomic coordinates from the traditionally refined X-ray structure of triclinic hen egg white lysozyme, resolved to 0.65… This particular system was chosen as an ensemble of 8 chemically realistic structures, which are used for the representation of observed structural variability within the crystallographic unit cell and which has been recently published [Falklöf et al. in Theor. Chem. Acc. 131:1076, 2012]. Optimizing only isolated residues within the protein for which all neighboring functional groups are fully identified, we show that in many cases dispersion-corrected DFT (and also Hartree-Fock) optimization competes with conventional refinement techniques. Significant correlations are found between method quality, perceived from small-molecule studies and changes in the R factor, indicating both the high quality of the original refinement but also indicating which methods will be most useful in subsequent full-protein refinements using imbedded DFT constraints.
Please use this identifier to cite or link to this item: