Computational protein crystallography : How to get the most out of your data

Sammanfattning: It is important to obtain accurate three dimensional structures of molecules and proteins to understand and predict their function and behaviour. X-ray crystallography is the major technique to determine three dimensional structures of proteins. Although there have been major improvements on the experimental side in determining crystallographic data, only small progress has been made on the computational side to get a correct model andinterpretation of this data.In small-molecule crystallography, some of the shortcomings in the model have already been overcome, but in protein crystallography they still remain. Therefore, we have adapted the Hirshfeld atom refinement from small-molecule crystallography to make it available also to protein crystallography. This enables improved modelling of high-resolution protein data. To achieve this goal, we combined the molecular fractionation with conjugate caps approach with the Hirshfeld atom refinement. We call this combined method fragHAR. With fragHAR, we could perform the first Hirshfeld atom refinement of a metalloprotein.Furthermore, we improved and applied the quantum refinement method, which employs quantum mechanics calculations to obtain a chemically and physically correct model for all parts of the protein, especially the active site. With quantum refinement, it is possible to distinguish between different interpretations of the structure, e.g. the elemental composition or the protonation state, even from medium-resolution crystallographic data. In this thesis, quantum refinement was improved for highly-charged systems by applying a continuum-solvent description of the surroundings in the quantum mechanics calculation. Furthermore, quantum refinement was applied to settle the nature of the unusual bidentate ligand in V-nitrogenase and the protonation state of the MoFe cluster in Mo-nitrogenase when inhibited by CO. For a recent structure of Mo-nitrogenase, we showed that there is no experimental support for the suggestion that N 2 is bound to the MoFe-cluster and presented a more likely model. We have also identified the most probable protonation states of the active site in acetylcholinesterase before and after inhibition by nerve agents. Finally, for triosephosphate isomerase we used a joint X-ray and neutron quantum refinement to investigate the hydrogen bond between an inhibitor and Lys-13.