Towards a better understanding of macromolecular X-ray structures

50多年来，X射线结晶学一直是确定生物大分子结构的主要方法。然而，原子结构并不是实验的直接结果：在测量和最终模型之间的大多数步骤中，都使用了大分子结构和实验方法的先验知识。因此，我们的结构取决于我们对其本质的基本理解--以及我们在结构模型和所采用的方法中表达这一点的能力。X射线数据和结构模型之间的差异通常以称为R值的百分比给出。小分子结构通常达到5%的R值，而大分子结构通常在20%-25%。处理后的数据和结构模型之间的这些相对较高的差异是一些相关的生物学问题无法回答的主要原因，例如配体是否结合，以及为什么一些结构根本无法解决，例如膜蛋白和大型复合体，这些结构通常只能获得低分辨率的数据。在大分子晶体中--与小分子晶体相反--通常有一半的体积被无序的溶剂占据，这会影响有序的大分子，使它们根据局部环境采用略有不同的构象。目前的模型并不能很好地描述这种疾病。我们使用模型阶段来生成映射的事实阻碍了进一步的改进。由于这引入了强烈的模型偏差，我们无法观察到我们没有预料到的东西。在这个项目中，从纯实验阶段计算的模型偏差自由映射将被用来克服这一挑战，并提供一个改进的大分子晶体结构模型。同时，我们将提高X射线数据的质量。最近开发的程序Auspex，它在结晶学数据中指示病理。目前，它主要用于冰环检测，但在本项目中将对其进行扩展，向用户展示数据采集和处理过程中的进一步问题，并提高我们对实际实验的理解。此外，我们还启动了Auspex网络服务器，方便结晶学家在线快速访问。有了这两个目标--改进的大分子晶体结构模型和更好的X射线数据--我们将能够降低R值，从而也能在新的和现有的结构中看到更多细节。这些改进将使世界各地的结构生物学家能够解决更具挑战性的生物学问题。

For over 50 years, X-ray crystallography has been the primary method to determine the structures of biological macromolecules. However, the atomic structure is not the direct outcome of the experiment: prior knowledge of macromolecular structures and experimental methods is used in most steps between measurement and the final model. Consequently, our structures are only as good as our fundamental understanding of their nature - and our ability to express this in structural models and the methods employed.The discrepancy between X-ray data and structural models is usually given as a percentage called the R value. While small molecule structures routinely reach R-values of 5%, macromolecular structures typically are at 20%-25%. These relatively high discrepancies between processed data and structural model are the main reason why some relevant biological questions - for exmple whether a ligand is bound - cannot be answered and why some structures cannot be solved at all, for example membrane proteins and large complexes, where often only low-resolution data are available.In macromolecular crystals - as opposed to small molecule crystals - typically half of the volume is occupied by disordered solvent, which affects the ordered macromolecules so that they adopt slightly different conformations dependent on the local environment. Current models do not describe this disorder very well. Further improvement is hindered by the fact that we use model phases to generate maps. Since this introduces a strong model bias, we cannot observe what we do not anticipate. In this project, model-bias free maps, calculated from purely experimental phases, will be used to overcome this challenge and to provide an improved model of macromolecular crystal structures. Parallel to this we will improve the quality of X-ray data. recently developed program, AUSPEX, which indicates pathologies in crystallographic data. Currently, it is mainly used to detect ice rings, but in this project it will be expanded to show users further problems during data collection and processing and improve our understanding of the actual experiment. In addition, we have initiated an AUSPEX web server for fast and easy access for crystallographers online.With these two objectives - improved structural models for macromolecular crystals and better X-ray data - we will be able to lower R-values and consequently, too see more detail in new as well as existing structures. These improvements will enable structural biologists worldwide to adress more challenging biological problems.