ABI Innovation: New Algorithms for Biological X-ray Free Electron Laser Data

结构生物学的重大进展往往源于数据收集和分析的新方法。从历史上看，越来越强大的x射线源导致了获取生物分子结构信息的能力的巨大飞跃，这与分子的生物学功能至关重要。第一台x射线自由电子激光器（XFEL）于2009年在SLAC国家加速器实验室上线，使研究人员能够前所未有地看到生物分子在自然状态（即温暖和潮湿）下的详细原子排列和运动。串行飞秒晶体学（SFX）是晶体学的一种新扩展，利用XFEL脉冲的极端通量和超短持续时间，对光合作用等基本重要过程产生了新的见解，有助于理解植物和细菌如何将阳光转化为能量并创造我们呼吸的氧气，使地球上的生命成为可能，并通过帮助药物设计来改善人类健康。例如，指导止痛药和血压调节器的改进。然而，针对生物靶点的XFEL实验通常需要非常大量的数据，因此需要相应大量的稀缺、珍贵的蛋白质，才能获得高分辨率的分子结构。此外，XFEL设备通常只能同时进行一到两个实验，目前只有两个XFEL服务于全球用户社区，而更多的XFEL将在2017年上线。可以研究的生物靶标的数量和发现的速度可以通过创新的发展，先进的算法，减少必要的测量数量和提取更多的信息，从样品中充分利用XFEL衍射独特的信息内容，并通过扩展XFEL使用到未结晶的目标显著增加。该项目有三个主要目标：(a)探索通过建模和优化提高XFEL系列飞秒晶体学数据精度的算法（超越数据合并的蒙特卡罗方法），(b)探索和开发利用XFEL二维和三维纳米晶体的全空间相干性的新相位方法，以及(c)开发可应用于无法结晶样品的结构测定方法，通过XFEL快速溶液散射（FSS）。FSS可以从粒子的“快照”衍射中提供动态结构信息，可以在溶液中动态研究，扩大了适用于XFELs的样品范围。这种方法的进一步发展，包括XFEL测量特有的统计强度相关性，可能完全取代晶体生长的需要。这些新算法将免费提供给科学界，通过显著减少获得高分辨率结构所需的样本数量、数据和实验时间（因此总体成本），将增加XFELs用于生物成像的革命性能力的可及性和适用性。因此，该项目将通过提供工作中的分子机器的时间分辨图像，直接有助于提高对基本生物分子机制的理解。该项目的结果将在http://www.public.asu.edu/~nzatsepi上公布

Major advances in structural biology have often resulted from novel approaches to data collection and analysis. Historically, increasingly powerful X-ray sources led to great leaps in the ability to obtain information about structures of biomolecules, which is critically related to the molecules' biological function. The first X-ray free electron laser (XFEL) came online in 2009 at SLAC National Accelerator Laboratory, granting researchers the unprecedented ability to see the detailed atomic arrangements in, and movements of, biological molecules in their natural state (i.e. warm and wet). Serial femtosecond crystallography (SFX), a novel extension of crystallography making use of the extreme flux and ultra-short duration of XFEL pulses, has produced new insights into fundamentally important processes such as photosynthesis, helping to understand how plants and bacteria convert sunlight into energy and create the oxygen we breathe, making life possible on earth, as well as improving human health by aiding pharmaceutical drug design, for example guiding improvements in pain killers and blood pressure regulators. However, XFEL experiments with biological targets typically require very large volumes of data and thus a correspondingly large volume of scarce, precious protein, to obtain high-resolution molecular structures. Furthermore, XFEL facilities typically host only one or two experiments simultaneously, and only two XFELs presently serve the global user community, while more will be coming online in 2017. The number of biological targets that can be studied and the rate of discoveries can be dramatically increased through the development of innovative, advanced algorithms that reduce the number of necessary measurements and extract more information from the samples by fully exploiting the information content that is unique to XFEL diffraction, and by extending XFEL use to uncrystallized targets. This project has three main objectives: (a) explore algorithms for improving data accuracy in XFEL serial femtosecond crystallography through modeling and optimization (beyond the Monte Carlo approach for data merging), (b) explore and develop novel phasing methods which exploit the full spatial coherence of the XFEL for 2D and 3D nanocrystals, and (c) develop structure-determination methods that can be applied to samples that cannot be crystallized, through XFEL fast solution scattering (FSS). FSS can provide dynamic structural information from "snapshot" diffraction from particles that can be studied dynamically in solution, broadening the range of samples suitable for XFELs. Further development of this approach to include statistical intensity correlations that are unique to XFEL measurements can potentially displace the need for crystal growth altogether. These new algorithms, which will be freely available to the scientific community, will increase accessibility and applicability of the revolutionary capabilities of XFELs for biological imaging by significantly decreasing the amount of sample, data and experimental time (and therefore overall costs) necessary to obtain high-resolution structures. Thus the project will contribute directly to improved understanding of the fundamental biomolecular mechanisms by providing time-resolved images of molecular machines at work. The results of this project will be available at http://www.public.asu.edu/~nzatsepi