MRI: Acquisition of a Cluster Computer for Digital Biology

该项目将基因组分析、结构蛋白质组学和大分子成像等领域的大量数据与生物功能的分子基础联系起来，从而影响到这些领域，旨在获得一个220个cpu的Opteron分布式内存计算集群。使用集群的结构与计算生物学和分子生物物理学（SCBMB）项目包括六个机构：贝勒医学院、莱斯大学、休斯顿大学、德克萨斯大学MD安德森分校、德克萨斯大学健康科学中心休斯顿分校和德克萨斯大学加尔维斯顿分校。该基础设施将服务于以下研究活动：国家大分子成像中心（NCMI）的生物组装体的电子低温显微镜重建（主要是电子低温显微镜和图像重建），结构蛋白质组学的功能表面表征（涉及进化痕迹方法（ET）和利用功能表面的高通量识别和几何匹配），比较基因组学（包括通过位置哈希（Pash）和比较序列组装和定位的DNA序列平行比较）微rna (miRNA)（通过调节基因表达结合靶互补mRNS）每个研究项目都有很高的计算需求，将满足所提出的并行环境。从原始生物数据中推导有意义的推论需要涉及数据集的密集计算。例如，为了重建大分子机器的三维图像，实验室需要从电子低温显微镜数据中转换和自动关联千兆字节的体素；为了确定蛋白质结构中的功能位点，实验室需要对数千种蛋白质结构进行全面比较；为了鉴定哺乳动物基因和检测新的微rna，实验室需要对数十亿个DNA序列的碱基对进行交叉比较。这些应用程序都具有重复但相对独立的计算的共同特征，这些计算可以在许多cpu之间进行拆分，并且不需要通过公共文件服务器进行通信。被请求的集群可以很好地服务这些共享特征。更广泛的影响：基础设施增强了参与机构的教育体验。SCBMB研究生课程（包括6所院校）的学生将使用该系统，大大增加了他们的研究机会。贝勒大学有面向本科生和高中生的拓展项目。讲习班、软件工具开发和技术转让将有助于传播结果。

This project, impacting genome analysis, structural proteomics, and macromolecular imaging by linking massive and exponentially growing data in each of these fields to the molecular basis of biological functions, aims at acquiring a 220-CPU Opteron distributed memory computing cluster. The Structural and Computational Biology and Molecular Biophysics (SCBMB) Program using the cluster encompasses six institutions: Baylor College of Medicine, Rice University, the University of Houston, the University of Texas MD Anderson, the University of Texas Health Sciences Center-Houston, and the University of Texas Medical Branch-Galveston. The infrastructure will service the following research activities:Electron Cryo-Microscopic Reconstruction of Biological Assemblies at the National Center for Macromolecular Imaging (NCMI) (mainly Electronic Cryomicroscopy and Image Reconstruction), Characterization of Functional Surfaces for Structural Proteomics (involving the Evolutionary Trace Method (ET) and utilizing High-Throughput Identification and Geometric Matching of Functional Surfaces), Comparative Genomics (including Parallel Comparison of DNA sequences via Positional Hashing (Pash) and Comparative Sequence Assembly and Mapping)Micro-RNA (miRNA) (binding target complementary mRNS by regulating gene expressions)Each research project has high computational demands that will be met with the proposed parallel environments. The derivation of meaningful inferences from raw biological data requires intensive computation involving data sets. For example, to reconstruct 3-D images of macromolecular machines, the labs need to transform and auto-correlate gigabytes of voxels from electron cryomicroscope data; to identify functional sites in protein structures, the lab requires all-against-all comparisons of thousands of protein structures; and to identify mammalian genes and detect novel micro-RNAs, the labs require cross-comparisons of billions of basepairs of DNA sequence. These applications all share the common feature of repeated but relatively independent computations that can be split among many CPUs with modest need of communication through a common file server. These shared characteristics can be serviced well by the requested cluster.Broader Impact: The infrastructure enhances the educational experience at participating institutions. Students in the SCBMB graduate program (encompassing 6 institutions) will use the system, significantly enhancing their research opportunities. Baylor has outreach programs for undergraduates and for high school. Workshops, software tool development, and technology transfer will serve to disseminate the results.