Parallel Software for Fast, Automated Determination of Virus Structures

描述(由申请人提供)：电子冷冻显微镜和基于计算机的3D图像重建技术正在彻底改变研究大型、复杂生物大分子机器结构的方式。这些方法为理解这些机器是如何工作的提供了关键。我们的目标是在最短的时间内以尽可能高的分辨率可靠地确定病毒结构。这将增强我们了解病毒如何导致各种疾病的能力，并可能为如何最好地开发抗病毒药物提供重要线索。显微镜和图像重建各自构成了一系列重大挑战，而新软件的开发显然是我们和其他人研究工作成功的关键。我们的目标是大幅提高计算能力，这些能力构成了我们低温重建工作中3D结构确定部分的核心。在低温重建中实现的分辨率和解决的结构受到许多因素的限制，其中最重要的包括采用的数值技术。因此，我们将专注于算法改进，以扩大可以研究的问题的范围，使其有可能达到更高的分辨率，并减少解决问题的时间。例如，我们将扩展软件的功能，以检查名义上的二十面体病毒的非二十面体组件以及对称性较低的粒子。除了上面提到的改进之外，我们的软件开发兴趣还包括自动化、并行计算和可用性的增强。对于许多病毒结构，特别是大型和/或不对称的病毒结构，图像重建是速率限制步骤。为了缩短图像获取和结构确定之间的时间，并缩小新手和专家用户可以达到的分辨率之间的差距，我们将增强AUTO3DEM，这是我们的自动化系统，用于智能集成图像重建所需的多种应用程序。该系统还提供了评估新想法的试验台，并使软件更容易为更广泛的用户所使用。拟议的工作应该使我们更接近我们的目标，即在显微镜下进行实时图像重建，或者至少与数据采集紧密结合，以提供快速反馈和快速筛选样品特性。我们的研究还受到与冷冻显微镜、图像处理、数据可视化和解释方面的同事的密切联系，在NIH、UCSD和Agouron基金会的支持下购买现代显微镜，以及与圣地亚哥超级计算机中心的互动。我们的所有软件都将便于电子显微镜社区访问。 公共卫生相关性：相关性病毒感染地球上几乎每一个活着的有机体，导致人类和其他动物的严重疾病和死亡，还极大地影响食物来源，往往导致饥荒和经济疲软。对病毒结构的了解为更好地了解病毒如何在易感宿主中自我复制提供了重要的联系。由电子显微镜记录的病毒图像生成的基于计算机的3D模型，提供了关于病毒分子结构的重要的、第一次一瞥。由于获得这些模型所需的时间通常需要几周或几个月的专心工作，因此这项提议的一个重要目标是将这一时间减少到几小时或几分钟。

DESCRIPTION (provided by applicant): Project Summary Electron cryo-microscopy and computer-based, 3D image reconstruction techniques are revolutionizing the way structures of large, complex biomacromolecular machines are studied. These methods provide keys to understanding how these machines function. Our goal is to determine virus structures reliably at the highest possible resolutions in the shortest amount of time. This will enhance our ability to understand how viruses cause a variety of dieases and may yield important clues about how best to develop anti-viral agents. Microscopy and image reconstruction each pose their own set of significant challenges, and development of novel software clearly represents a key to the success of our research efforts as well as that of others. We aim to substantially enhance the computational capabilities that comprise the heart of the 3D structure determination part of our cryo-reconstruction work. The resolutions achieved and structures solved in cryo-reconstruction are limited by many factors, not the least of which includes the numerical techniques employed. Hence, we will focus on algorithmic improvements that expand the range of problems that can be studied, make it possible to reach higher resolutions, and reduce time to solution. For example, we will extend the capabilities of our software to examine the non-icosahedral components of nominally icosahedral viruses as well as particles with lower symmetries. In addition to the improvements noted above, our software development interests include automation, parallel computing, and enhancements to usability. For many virus structures, especially large and/or asymmetric ones, image reconstruction is the rate limiting step. To reduce the time between image acquisition and structure determination and also narrow the gap between the resolutions that can be reached by novice and expert users, we will enhance AUTO3DEM, our automation system for intelligently integrating the multiple applications required in image reconstruction. This system also provides a test bed for evaluating new ideas and makes the software easily accessible to a wider range of users. The proposed work should take us closer to our goal of real-time image reconstruction at the microscope, or at least tightly coupled to data acquisition, to provide rapid feedback and quickly screen sample properties. Our studies are also heavily leveraged by close ties to colleagues in cryo-microscopy, image processing, and data visualization and interpretation, the acquisition of modern microscopes with support from NIH, UCSD, and the Agouron foundation, and interaction with the San Diego Supercomputer Center. All of our software will be made readily accessible to the electron microscopy community. PUBLIC HEALTH RELEVANCE: Relevance Viruses infect nearly every living organism on Earth, causing serious diseases and death in humans and other animals, and also greatly impact food sources often leading to famine as well as weakened economies. Knowledge of virus structure provides an important link to better understand how viruses replicate themselves in susceptible hosts. Computer-based 3D models generated from virus images recorded in transmission electron microscopes, provide an important, first glimpse about the molecular structures of viruses. As the time required to obtain these models often requires weeks or months of dedicated effort, an important goal of this proposal is to reduce this down to hours or minutes.