The comparable order of magnitude between interatomic distances in a crystal and the wavelength of X-rays make X-ray crystallography the ideal analytical tool to gain insight into the structure of crystalline material, including biomolecules. Nevertheless, biomolecular crystallography has until now relied on the successful growth of single crystals of suitable size and quality. These remain the exception rather than the rule, since biomolecules often produce polycrystalline precipitate instead. Yet, an interest in making use of the once-discarded polycrystalline material, through the technique of powder diffraction, has only recently emerged. This can be accounted for by the information deficit which powder diffraction data suffers from in comparison with that of single crystal. The paucity of information in powder data stems from the compression of the three-dimensional reciprocal space onto the one-dimension of a powder pattern. In spite of this, powder diffraction holds the potential for application in biomolecular crystallography as is shown in the two different studies presented herein. Both studies were carried out with methods which do not rely on employing previously determined crystal-structures as molecular models. This therefore allowed the objective assessment of the quality of information that powder diffraction data can contribute to the structural investigation of biomolecules. In the first project, the traditional single-crystal structure-solution process is applied to data extracted from protein powder diffraction patterns measured on a synchrotron source. The use of models is avoided by employing the de novo phasing method of isomorphous replacement. With two protein test-cases, namely hen egg white lysozyme and porcine pancreatic elastase, it is demonstrated that protein powder diffraction data can afford structural information up to medium resolution. Indeed, a single isomorphous replacement analysis generated molecular envelopes accurately describing the crystal packing of both protein systems, while a multiple isomorphous replacement experiment, carried out only on lysozyme, revealed an electron density map in which elements of the secondary structure could be located. In fact, the resolution of the latter was discovered to be sufficient to determine the chirality of the protein molecule it represented. In addition to being encouraging, these results do not reflect the full potential of biomolecular powder diffraction, due, in large part, to the ultimately unsuitable nature of one type of phasing method, single crystal, being applied to another type of data, powder. An alternative approach to extract information from protein powder diffraction data is to employ powder-specific structure-solution techniques, such as global optimization methods. Although these methods make use of a starting "model", it is a molecular description of the system under study based on known chemical quantities rather than a related molecular configuration based on a previously determined crystal structure. Since their conception, global optimization methods have continuously been developed to enable the tackling of increasingly complex crystal structures. However, the immense complexity of biomacromolecules has kept proteins well out of reach of such methods. In an attempt to further reduce the gap separating the two levels of complexity, the second study reported herein puts forth the implementation of Ramachandran plot restraints into the algorithm of a global optimization method, i.e. that of simulated annealing. More specifically, the Ramachandran plot was approximated using a two-dimensional Fourier series which was subsequently expressed as a penalty function and incorporated into the search algorithm.
EPFL2011
