Acquisition of a Parallel Computer for Structural Biology

这项提议寻求为一种新的高性能并行计算机寻求资金，用于计算生物学和化学，特别是研究生物大分子的结构和功能。华盛顿州立大学在生物大分子研究方面已经建立了相当大的实力，其中核磁共振中心和新的X射线结晶学设施都是重要的组成部分。然而，缺少的关键功能是缺乏计算能力。计算机提供了关键的环节，因为它们既需要处理从核磁共振和晶体结构确定实验中获得的大量数据集，也需要用于理论研究，如分子动力学模拟和电子结构计算，这些研究提供了结构和功能之间的联系，也为实验提供了新的方向。结构生物学的现代研究需要快速的计算机，这是因为实验方法的快速发展需要分析越来越大的数据集，而且计算机模拟在理解结构功能关系方面的重要性越来越大。除了日常计算需求外，五名主要调查人员中的每一位都有需要大规模计算能力的项目。这些项目包括对肌球蛋白、酶底物复合体、电子转移复合体、蛋白质-核酸复合体以及水在蛋白质功能中的作用的研究。目前，华盛顿州立大学的结构生物学计算是在个人研究人员拥有的工作站上进行的。这些工作站中的许多已有5年以上的历史，这意味着它们处于硬件可靠性的极限，在速度上比当前的工作站落后几个数量级。例如，目前正在进行的计算机模拟需要在我们最快的工作站(IBM RS/6000 350)上花费数周的计算机时间。唯一的大学设施是一台IBM 3090-300和一台VAX集群(比许多个人电脑还慢)，它们严重过度使用和昂贵。此外，我们最常用的商业软件不适用于这些计算机。虽然新的廉价工作站(比我们的RS/6000快5到7倍)可以用来解决许多简单的问题，但华盛顿大学的研究人员面临着越来越多的问题，这些问题需要非常快的计算、巨大的内存和/或巨大的磁盘空间，这些问题超出了个人研究人员的能力范围，即使是那些使用计算机作为主要研究工具的研究人员也是如此。这些需求部分可以通过超级计算中心来满足，但这些中心正在变得超负荷，因此该路线和工作站之间的中间水平是该园区目前无法满足的重要需求。像这里提出的这样的并行处理计算机是学术研究的理想解决方案。它们的速度可以与克雷计算机等向量超级计算机相当，但成本只有它的一小部分。这些计算机具有速度最快的工作站中的多个相同类型的处理器，但将这些处理器组合到一个机箱中比多个独立的工作站具有几个重要的优势。首先，通过将计算“负载”分配到几个处理器上，计算速度几乎可以随着处理器数量的增加而线性增加。其次，昂贵的资源，如内存和软件，可以由多个用户共享。第三，单台机器的操作和维护少得多，大大节省了人员时间。第四，不同的处理器可以在不同的时间专用于不同类型的使用，从而允许高效地使用处理器。最后，模块化的本质允许调查人员在资金充足的情况下，通过简单地增加额外的处理器来提高机器的效率。拟议的计算机将取代我们过时的工作站，以满足日常计算需求，但更重要的是，将允许WSU的调查人员探索在这些工作站上不可能实现的新途径。此外，它不仅将对目前直接参与结构生物学的研究人员产生影响，还将提供一种手段，向整个校园的研究人员介绍结构生物学和化学中最先进的计算方法。任何联网的个人计算机或工作站都可以访问这台计算机，并且已经有提供咨询的工作人员。随着新教员的增加和现有教员对结构-功能研究的兴趣日益浓厚，获得一台新的计算机至关重要。

This proposal seeks funding for a new high performance, parallel computer for computational biology and chemistry, specifically to study the structure and function of biological macromolecules. Washington State University has built considerable strengths in the studies of biological macromolecules, of which both the Nuclear Magnetic Resonance (NMR) Center and the new X-Ray Crystallographic facilities are important components. However, the key feature missing is the lack of computing power. Computers provide the crucial link since they are needed both for processing the large data sets obtained from NMR and crystallographic structure determination experiments as well as for theoretical studies such as molecular dynamics simulations and electronic structure calculations that provide the connection between structure and function and that also provide new directions for experiments. Modern research in structural biology requires fast computers, with ever increasing needs because of rapid advances in experimental methods that require analysis of larger and larger data sets and because of the increasing importance of computer simulations in understanding structure function relationships. In addition to daily computational needs, each of the five principal investigators has projects that require massive computing capabilities. These projects include studies of myosin, enzymesubstrate complexes, electron transfer complexes, protein-nucleic acid complexes, and the role of water in protein function. Currently, computing for structural biology at WSU is being carried on workstations owned by individual investigators. Many of these workstations are more than 5 years old, which means that they are at the limit of hardware reliability and are orders of magnitude behind current workstations in speed. For instance, computer simulations currently being carried out take weeks of computer time on our fastest workstations (IBM RS/6000 350). The only university facilities are an IBM 3090-300 and a VAX cluster (slower than many personal computers), which are heavily overused and expensive. Moreover, our most commonly used commercial software is not available for these computers. Although new inexpensive workstations (5 to 7 times faster than our RS/6000) may be used for many simple problems, researchers at WSU are facing more and more problems that require very fast computation, huge amounts of memory and/or huge amounts of disk space beyond the reach of individual investigators, even those that use computers as a major research tool. These needs can be met in part by Supercomputing Centers, but these centers are becoming overloaded so a middle level between that route and workstations is a vital need not currently met on this campus. Parallel processing computers such as the one proposed here are an ideal solution for academic research. They can have speeds equivalent to vector supercomputers such as Cray computers, but at a fraction of the cost. These computers have multiple processors of the same type that are found in the fastest workstations, but the combination of these processors into a single chassis provides several important advantages over multiple independent workstations. First, computation speed can increase almost linearly with the number of processors by distributing the computational "load" across several processors. Second, expensive resources such as memory and software can be shared by multiple users. Third, operations and maintenance of a single machine are much less, leading to a great saving in personnel time. Fourth, different processors may be dedicated to different types of usage at different times, allowing efficient usage of the processors. Finally, the modular nature allows investigators to increase the efficiency of the machine by simply adding additional processors if money becomes available. The proposed computer will replace our outdated workstations in fulfilling routine computational needs, but more importantly, will allow investigators at WSU to pursue the new avenues not possible on those workstations. Moreover, it will have an impact on not only the researchers currently directly involved in structural biology, but will provide a means of introducing state-of-the-art computational methods in structural biology and chemistry to researchers across the campus. The computer will be accessible by any networked personal computer or workstation, and staff are already in place who would provide consultation. With the addition of new faculty and the growing interest of existing faculty in structure-function studies, it is crucial that a new computer be obtained.