OAC Core: High Performance Computing Algorithms and Software for large-scale Mass Spectrometry based Omics

基于高维质谱(MS)的组学可以对数千种蛋白质进行系统分析，有望发现各种疾病的新生物标记物，并更好地了解人类系统生物学。元蛋白质组学也是研究不同环境中微生物的基础，对人类、农业、水生、陆地、能源和大气系统都有重要影响。除了质谱学硬件的进步外，对这些高通量质谱仪产生的复杂数据进行有效和可扩展的分析还需要越来越复杂的计算工具。该项目将设计和开发高性能计算框架，从而能够有效地分析质谱机产生的组学数据。拟议的高性能计算(HPC)技术将能够识别新的多肽/蛋白质，并深入了解微生物群落及其对人类健康、农业和环境的影响。拟议的研究和教学活动将向学生介绍高性能计算、大数据计算生物学和数据密集型计算。这项拟议的工作还将培训博士生，包括西班牙裔服务机构的博士生。质谱机产生的数百万个光谱与用于肽推断的万亿级理论数据库进行比较。目前，这一推论的大部分是使用串行算法完成的，对于大型数据库来说，这可能需要数周的计算时间。本研究的主要技术目标是为各种异类体系结构设计、开发和评估高性能计算(HPC)基础设施。这种可在异类架构上运行的避免通信的HPC算法将使MS数据匹配的非模型万亿级数据库能够进行可伸缩的MS数据分析，这是目前难以逾越的技术障碍。该项目将致力于以下几个方面的设计和开发：(1)基于CPU-GPU的大规模MS组学处理方法和避免通信的并行流水线；(2)在单个节点上开发多个GPU的方法，该方法扩展到超级计算机器上的内存分布式CPU-GPU节点；(3)基于CPU-FPGA架构的硬件/软件协同设计。这种计算基础设施将允许科学家使用大型异质超级计算机，硬件/软件设计的开发将使我们能够将半导体设计直接整合到质谱仪上。为终端应用(MS组学数据)设计的此类半导体的开发将保持美国的全球经济竞争力，并将加速迫切需要的个性化营养研究、人体肠道微生物组研究和癌症治疗研究。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

High dimensional Mass Spectrometry (MS) based omics allows systematic analysis of thousands of proteins with the promise of discovering new biomarkers for various disease conditions and better understanding of human systems biology. Meta-proteomics is also fundamental to studies of microorganisms in diverse environments, and has significant effect on human, agriculture, aquatic, terrestrial, energy, and atmospheric systems. In addition to advances in mass spectrometry hardware, effective and scalable analysis of the complex data produced by these high-throughput mass spectrometers requires increasingly sophisticated computational tools. This project will design and develop high-performance computational frameworks which will enable effective analysis of omics data produced from mass spectrometry machines. The proposed high-performance computing (HPC) techniques will enable identification of novel peptides/proteins, and insights into microbiome communities and their effects on human health, agriculture, and environments. The proposed research and teaching activities will introduce students to high-performance computing, big data computational biology, and data-intensive computing. The proposed work will also train PhD students, including those at a Hispanic Serving Institution.Millions of spectra generated from mass spectrometry machines are compared with tera-scale theoretical database for peptide deductions. Currently, the bulk of this deduction is accomplished using serial algorithms which may take weeks of computational time for large databases. The overarching technical objective of this study is to design, develop, and evaluate high performance computing (HPC) infrastructure for a variety of heterogenous architectures. Such communication-avoiding HPC algorithms that can run on heterogenous architectures will enable scalable MS data analysis for non-model tera-scale databases against which MS data is matched and is currently an insurmountable technical hurdle. The project will focus on the design and development of: (1) CPU-GPU based method for processing of large-scale MS-based omics with communication-avoiding parallel pipelines; (2) methods for exploiting multiple GPUs on single node which is extended to memory-distributed CPU-GPU nodes on supercomputing machines; (3) hardware/software co-designs using CPU-FPGA architectures. This computational infrastructure will allow scientists to use large heterogeneous supercomputers, and the development of hardware/software designs will enable us to incorporate semiconductor designs directly on mass spectrometry machines. The development of such semiconductors designed for end-use application (of MS omics data) will preserve US global economic competitiveness, and will accelerate urgently needed personalized nutrition studies, human gut microbiome research, and cancer therapeutics studies.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.