Automated, optimized, intelligent data collection for cryo-EM

冷冻电镜的自动化、优化、智能数据采集

• 批准号: 10317907
• 负责人: Gabriel C Lander
• 金额: $ 65.5万
• 依托单位: SCRIPPS RESEARCH INSTITUTE, THE
• 依托单位国家: 美国
• 财政年份: 2021
• 资助国家: 美国
• 起止时间: 2021-09-22 至 2025-06-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10317907
• 关键词: 2019-nCoV; Adoption; Algorithms; Automation; Automobile Driving; Biophysics; Buffers; Bypass; COVID-19; Code; Communities; Computer software; Country; Cryoelectron Microscopy; Data; Data Collection; Data Set; Databases; Development; Disease; Electron Microscope; Electron Microscopy; Electrons; Feedback; Foundations; Generations; Goals; Hand; Heterogeneity; Ice; Image; Image Analysis; Institution; Intelligence; Intervention; Learning; Maintenance; Manuals; Methodology; Methods; Microscope; Molecular; Molecular Conformation; Output; Particle Size; Pathologic; Planets; Play; Preparation; Process; Research; Resolution; Role; Running; Sampling; Series; Site; Specimen; Structure; Thick; Time; Training; United States National Institutes of Health; Update; Vaccines; Virus; Work; application programming interface; base; biological systems; cluster computing; comparative; computer infrastructure; computer program; computerized data processing; convolutional neural network; data acquisition; data mining; data quality; design; fighting; image processing; improved; innovation; migration; neutralizing antibody; next generation; open source; particle; portability; prediction algorithm; preservation; programs; real-time images; reconstruction; structural biology; therapeutic development; tool

Project Summary Cryo-electron microscopy (cryo-EM) is now a widely established and indispensable method for determining the high-resolution structures of biomedically important molecules. Given that thousands of images, often acquired over the course of several days, are required to obtain such structures, automation software has played a critical role in the large-scale adoption of this method by the scientific community. In just the past five years, cryo-EM has revolutionized our understanding of entire biological systems, and in 2020 provided the first molecular descriptions of SARS-CoV-2 interaction with neutralizing antibodies. The widespread adoption of cryo-EM recently prompted the NIH to invest in three National Centers through the Transformative High Resolution Cryo- Electron Microscopy Program, providing free, high-end electron microscope access to biologists across the country. The exponential increase in the popularity of cryo-EM has led to an astonishing number of developments in sample preparation methodologies and image processing algorithms, which have improved attainable resolution of single particle reconstructions. However, comparatively little progress has been made in optimizing the quality of the cryo-EM data being collected. The pioneering software packages Leginon and Appion demonstrated the power of automated data acquisition and real-time processing (respectively), and there are now numerous programs for automated data acquisition and real-time processing. Despite advances in automation, optimally extracting the highest quality data from an EM sample still requires manual involvement of an expert electron microscopist. User intervention and expertise is necessary to run the appropriate image analyses, interpret the results, and make informed decisions on how the processed results relate to the ongoing data collection. However, even experts must content with the fact that the “best grid regions” differ drastically from sample to sample, and there are no established tools for automatically and quickly assessing the quality of the specimen across the various microenvironments of an EM grid. Given the ever-increasing incorporation of cryo-EM into labs’ research programs, it is imperative that data collection and processing be streamlined to match the growing needs of the structural community. We propose to develop a second generation Leginon/Appion software package, “Magellon”, to overcome existing bottlenecks and provide an avenue toward fully automated data acquisition that bypasses need for user input during data collection. Importantly, this software will support the computational infrastructure to enable real-time image processing results to inform on and modify the ongoing data collection regime by learning where to acquire images in regions that will yield the highest resolution structures. We will develop and incorporate new, fast image assessment routines, while also providing an application programming interface to enable the incorporation of extensions and plugins from developers in the community. Further, Magellon will enable straightforward, seamless import and export of data from its database to accommodate remote data acquisition at any of the regional or national cryo-EM centers.

项目摘要 冷冻电子显微镜（cryo-EM）现在是一种广泛建立的和必不可少的方法，用于确定 生物医学重要分子的高分辨率结构。鉴于数千张图像，通常是在 在几天的过程中，需要获得这样的结构，自动化软件发挥了关键作用。 在科学界大规模采用这种方法方面发挥了重要作用。仅仅在过去的五年里， 彻底改变了我们对整个生物系统的理解，并在2020年提供了第一个分子 描述SARS-CoV-2与中和抗体的相互作用。冷冻电镜的广泛应用 最近促使美国国立卫生研究院通过变革性高分辨率冷冻技术投资三个国家中心， 电子显微镜计划，提供免费的，高端的电子显微镜访问整个生物学家 国家低温EM的普及率呈指数级增长，导致了惊人数量的发展 在样品制备方法学和图像处理算法中， 单粒子重建的分辨率。然而，在优化方面取得的进展相对较小 所收集的冷冻电镜数据的质量。开创性的软件包Leginon和Appion 展示了自动数据采集和实时处理的能力（分别）， 现在有许多程序用于自动数据采集和实时处理。尽管取得了进展， 自动化，从EM样本中最佳提取最高质量的数据仍然需要人工参与， 电子显微镜专家运行适当的映像需要用户干预和专业知识 分析，解释结果，并就处理结果如何与正在进行的 数据收集。然而，即使是专家也必须满足于“最佳网格区域”差异巨大的事实 从一个样本到另一个样本，没有既定的工具来自动和快速评估质量， 在EM网格的各种微环境中的样本。鉴于越来越多地纳入 冷冻EM进入实验室的研究计划，必须简化数据收集和处理， 满足结构化社区日益增长的需求。我们建议开发第二代 Leginon/Appion软件包“Magellon”，以克服现有的瓶颈， 完全自动化的数据采集，在数据采集过程中无需用户输入。重要的是这 软件将支持计算基础设施，使实时图像处理结果能够提供信息， 并通过学习在哪里获取将产生 最高分辨率结构。我们将开发和整合新的快速图像评估程序，同时还 提供应用程序编程接口以使得能够结合来自 开发商在社区。此外，Magellon还将实现简单、无缝的数据导入和导出 从其数据库，以适应远程数据采集在任何区域或国家冷冻EM中心。