APPLICATION OF HIGH-PERFORMANCE COMPUTING TO THE RECONSTRUCTION, ANALYSIS, AND

高性能计算在重建、分析和预测中的应用

• 批准号: 8364313
• 负责人: RICK L. STEVENS
• 金额: $ 0.11万
• 依托单位: CARNEGIE-MELLON UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-09-15 至 2013-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/8364313
• 关键词: Algorithms; Animal Model; Anti-Bacterial Agents; Area; Biochemical Pathway; Biology; Biomedical Research; Chemicals; Collection; Communities; Computer software; Conditioned Culture Media; Custom; Data; Data Collection; Engineering; Environment; Equilibrium; Flow Cytometry; Funding; Future; Genes; Genome; Grant; Growth; High Performance Computing; Hour; Knock-out; Manuals; Metabolic; Methods; Microfluidic Microchips; Modeling; National Center for Research Resources; Organism; Phenotype; Principal Investigator; Process; Prokaryotic Cells; Publishing; Research; Research Infrastructure; Resources; Shotgun Sequencing; Simulate; Source; Technology; United States National Institutes of Health; Update; computing resources; cost; drug development; genome sequencing; high throughput technology; microorganism; open source; reconstruction; simulation

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. The field of biology is undergoing a fundamental shift from a data poor field to a data rich field thanks to the advent of numerous new high-throughput technologies for the collection of experimental data: Shotgun sequencing, Pyrosequencing, Microarrays, ChIP-chip, Biolog phenotyping arrays, microfluidic devices, and flow cytometry. Today simulation is struggling to keep pace with data collection in many areas, and nowhere is this more evident than in genome-sequencing versus genome-scale modeling. While over 800 prokaryotic genomes have been sequenced in the past ten years, only 30 genome-scale metabolic models have been published, and the pace of genome-sequencing continues to increase threatening to extend this already massive gap. Today, the paradigm of the genome-scale metabolic modeling community is that it requires a year or more of manual effort to produce a new model of a microorganism. However, technologies have emerged in recent years that make it possible to automate or expedite various steps of the genome-scale reconstruction process, and we have recently tied these technologies together into an automated genome-scale model reconstruction pipeline. While this pipeline makes it possible to construct a single model in one to five days, extensive computation is required in this reconstruction process. We are proposing to use the computational resources in the TerraGrid to apply this reconstruction process to build new genome-scale metabolic models for every prokaryote with a completely sequenced genome. We then plan to use these models in a number of high-impact scientific studies including: (1) simulating the knockout of every metabolic gene to study the robustness of the metabolic networks of these organisms and identify new potential targets for future antibacterial drug development, (2) simulating growth of each microorganism in a variety of chemical conditions to identify the environments in which various communities of microorganisms are capable of surviving, (3) predicting the minimal defined media conditions that are required in order to culture each modeled organism, and (4) simulating the engineering of each modeled organism to produce organic compounds of industrial value from a variety of renewable raw materials. While these studies produce very different results and appeal to fundamentally different application areas, they all can be accomplished by applying the Flux Balance Analysis method to the genome-scale metabolic models we will be constructing. The primary algorithm we will be using in the proposed reconstruction and analysis of genome-scale models is flux balance analysis. The most significant computation involved in this algorithm is the solving of a linear or mixed integer linear optimization problem. Fortunately, numerous open source software is available for solving linear and mixed integer linear optimization problems. We will be applying the GLPK, SCIP, and BCP solvers along with our own custom built MPI-ready FBA software to perform all of the proposed reconstruction and analysis calculations. In total, we expect that 104 distinct mixed integer linear optimization problems and 1012 distinct linear optimization problems will need to be solved in the first year of this project, requiring a total of 1.5 million CPU hours. In the two following years, we anticipate an equal number of simulations will be required due to the release of additional sequenced organism and the update of the annotations in the existing organisms.

这个子项目是许多利用资源的研究子项目之一 由NIH/NCRR资助的中心拨款提供。子项目的主要支持 而子项目的主要调查员可能是由其他来源提供的， 包括其它NIH来源。 列出的子项目总成本可能 代表子项目使用的中心基础设施的估计数量， 而不是由NCRR赠款提供给子项目或子项目工作人员的直接资金。 生物学领域正在经历从数据贫乏领域到数据丰富领域的根本转变，这要归功于许多用于收集实验数据的新的高通量技术的出现：鸟枪测序、焦磷酸测序、微阵列、ChIP芯片、Biolog表型分析阵列、微流体装置和流式细胞术。今天，在许多领域，模拟都在努力跟上数据收集的步伐，这一点在基因组测序与基因组规模建模中表现得最为明显。虽然在过去的十年中已经测序了800多个原核生物基因组，但只有30个基因组规模的代谢模型已经发表，并且基因组测序的步伐继续增加，可能会扩大这一已经巨大的差距。今天，基因组规模的代谢建模社区的范例是，它需要一年或更长时间的人工努力来产生微生物的新模型。然而，近年来出现了一些技术，可以自动化或加快基因组规模重建过程的各个步骤，我们最近将这些技术结合在一起，形成了一个自动化的基因组规模模型重建管道。虽然这个管道可以在一到五天内构建一个模型，但在这个重建过程中需要大量的计算。我们建议使用TerraGrid中的计算资源来应用这个重建过程，为每个具有完整测序基因组的原核生物建立新的基因组规模的代谢模型。然后，我们计划在一些高影响力的科学研究中使用这些模型，包括：（1）模拟每个代谢基因的敲除，以研究这些生物体的代谢网络的鲁棒性，并鉴定未来抗菌药物开发的新的潜在靶标，（二）模拟每种微生物在各种化学条件下的生长，以鉴定各种微生物群落所处的环境，能够存活，（3）预测培养每种模拟生物体所需的最低限定培养基条件，和（4）模拟每种模拟生物体的工程化，以从各种可再生原料生产具有工业价值的有机化合物。虽然这些研究产生了非常不同的结果，并呼吁从根本上不同的应用领域，他们都可以通过应用通量平衡分析方法，我们将构建基因组规模的代谢模型来完成。我们将在所提出的基因组规模模型的重建和分析中使用的主要算法是通量平衡分析。该算法涉及的最重要的计算是线性或混合整数线性优化问题的求解。幸运的是，许多开源软件可用于解决线性和混合整数线性优化问题。我们将应用GLPK、SCIP和BCP求解器沿着以及我们自己定制的MPI就绪FBA软件来执行所有建议的重建和分析计算。总的来说，我们预计在这个项目的第一年需要解决104个不同的混合整数线性优化问题和1012个不同的线性优化问题，总共需要150万个CPU小时。在接下来的两年里，由于额外测序生物的发布和现有生物注释的更新，我们预计将需要相同数量的模拟。