Novel algorithms and hardware designs for ultra-fast next-gen sequence analysis

用于超快速下一代序列分析的新颖算法和硬件设计

• 批准号: 8025626
• 负责人: Onur Mutlu
• 金额: $ 36.93万
• 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: 2011-06-20 至 2015-04-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/8025626
• 关键词: Address; Algorithms; Architecture; Area; Atlases; Computer Architectures; Computer software; Consumption; DNA Sequence; Data; Data Analyses; Dependence; Development; Disease Pathway; Engineering; Gene Expression Profile; Genes; Genetic; Genome; Genomics; Goals; Healthcare Systems; Human Genome; Individual; Institution; Knowledge; Large-Scale Sequencing; Length; Location; Maintenance; Maps; Medical; Medical Genetics; Medicine; Memory; Metagenomics; Methods; Methylation; Modeling; Pattern; Performance; Pharmacogenomics; Polymorphism Analysis; Preventive; Process; Property; RNA analysis; Reading; Reporting; Research; Research Infrastructure; Running; Sequence Analysis; Small RNA; Software Tools; Source; Speed; Stream; System; Techniques; Technology; Time; Translations; Variant; Work; abstracting; base; cluster computing; comparative; computer infrastructure; computing resources; cost; design; epigenetic variation; genome sequencing; histone modification; improved; instrument; mutant; new technology; next generation; novel; parallel computing; processing speed; prototype; software development; success; tool; tumor

DESCRIPTION (provided by applicant): Abstract With the introduction of next generation sequencing (NGS) technologies, we are facing an exponential increase in the amount of genomic sequence data. These new methods have already started to fundamentally revolu- tionize the area of genome research through low-cost and high-throughput genome sequencing. NGS technologies promise to impact a broad range of genetic applications. These include, but are not limited to, large-scale sequencing studies, polymorphism detection, small RNA analysis, metagenomics, com- parative genomics, discovery of epigenetic variation (histone modification and methylation patterns), charac- terization of tumor DNA sequences, identification of mutant genes in disease pathways and transcriptome profiling. Low-cost sequencing will impact the whole health care system because sequencing of personal genomes will be a part of preventive and personalized medicine as a result of potential advancements in phar- macogenomics. The overall data throughput generated by these new technologies is enormous: for example, in the Illumina Genome Analyzer, each run produces up to 1 billion reads and >100 Gb of basepairs of sequence data. Due to the lower cost of these methods, large genome centers have started to upgrade their sequencing capa- bilities, and are now able to generate 500 gigabases of data per day when 40 instruments are used. Such large amounts of data overwhelm existing computational resources, and urgent action is needed to enable the translation of this rich new source of genomic information into medical benefit. The success of all medical and genetic applications of next-generation sequencing critically depends on the existence of computational tech- nologies that can process and analyze the enormous amounts of sequence data fast and in an energy-efficient manner. The goal of this proposal is to develop such technologies by combining the benefits of enhanced software algorithms and specialized hardware accelerators. Our proposed research aims to accelerate next generation sequence analysis 1000-fold or more by combin- ing our knowledge in genomic sequence analysis, algorithms development, and computer architecture/engineering. Our plan to address the problems of processing unprecedented amounts of sequence data has three major components. First, we will develop and improve sophisticated software algorithms and tools to handle large amounts of sequence reads generated by all major NGS platforms without sacrificing sensitivity while cor- recting for the sequencing biases associated by each of the NGS platforms. Our algorithms will also be able to map reads in the duplicated regions of the genome and report the underlying sequence variation, an important feature especially to characterize segmental duplications and structural variation that no other read mapping tool can currently achieve. Second, we will boost the performance and efficiency of our algorithms (100 to 1000-fold) by accelerating the required inherently-parallel computations of the sequence search problem on massively-parallel hardware engines available today, graphics processing units (GPUs). Finally, we will design specialized hardware architectures to enhance the speed of sequence analysis beyond orders of magnitude while reducing energy consumed by it by 100-fold or more. Our research will broadly impact large-scale genome studies such as the 1000 Genomes Project, the Can- cer Genome Atlas Project, and the ENCODE Project, by not only increasing their ability to reach conclusions very fast but also reducing their energy consumption and maintenance costs related to maintaining compu- tation clusters for data analysis. Our research, if successful, can eliminate the dependence of sequence analysis on large and power-hungry computing clusters/data-centers, thereby making sequence analysis significantly cheaper and energy-efficient, and hence enabling sequence analysis to be performed by the main- stream without the need to build large computational infrastructures. Together with further advances in sequencing technologies, research resulting from this proposal can help personal genomics become a reality: advancement and application of pharmacogenomics will start the era of personalized medicine. Through ultra- fast, energy-efficient and cost-efficient sequence analysis, this study can pave the way to unlimited number of new discoveries by making it feasible to analyze terabases of sequence data that cannot currently be handled with existing computational processing power. PUBLIC HEALTH RELEVANCE: Next-generation sequencing (NGS) technologies promise the era of preventive and personalized medicine through low-cost and high-throughput genome sequencing. The success of all medical and genetic applica- tions of next-generation sequencing critically depends on the existence of computational technologies that can process and analyze the enormous amounts of sequence data fast and in an energy-efficient manner without requiring the building of large infrastructures. The goal of this proposal is to develop such technologies by combining the benefits of enhanced software algorithms and specialized hardware accelerators.

描述(由申请人提供)： 摘要随着下一代测序技术的引入，我们面临着基因组序列数据量的指数增长。这些新方法已经开始通过低成本和高通量的基因组测序从根本上改变基因组研究领域。NGS技术有望对广泛的遗传应用产生影响。这些包括但不限于大规模测序研究、多态检测、小RNA分析、元基因组学、比较基因组学、表观遗传变异的发现(组蛋白修饰和甲基化模式)、肿瘤DNA序列的特征、疾病途径中突变基因的鉴定和转录组图谱。低成本的测序将影响整个医疗保健系统，因为由于药物宏基因组学的潜在进步，个人基因组测序将成为预防和个性化医学的一部分。这些新技术产生的总体数据吞吐量是巨大的：例如，在Illumina基因组分析仪中，每次运行产生高达10亿次读取和100 GB的基对序列数据。由于这些方法的成本较低，大型基因组中心已经开始升级其测序能力，现在使用40台仪器时，每天能够产生500G的数据。如此大量的数据淹没了现有的计算资源，迫切需要采取行动，将这种丰富的新基因组信息源转化为医疗效益。下一代测序的所有医学和遗传学应用的成功关键依赖于能够快速、节能地处理和分析海量序列数据的计算技术的存在。这项提议的目标是通过结合增强的软件算法和专门的硬件加速器的好处来开发这种技术。我们提出的研究旨在通过结合我们在基因组序列分析、算法开发和计算机体系结构/工程方面的知识，将下一代序列分析速度提高1000倍或更多。我们解决处理史无前例数量的序列数据的问题的计划有三个主要组成部分。首先，我们将开发和改进复杂的软件算法和工具，以处理所有主要NGS平台产生的大量序列读取，而不牺牲灵敏度，同时纠正每个NGS平台相关的测序偏差。我们的算法还将能够映射基因组复制区域中的阅读并报告潜在的序列变异，这是一个重要的功能，特别是对于表征片段复制和结构变异，这是目前任何其他RED映射工具都无法实现的。其次，我们将通过在目前可用的大规模并行硬件引擎-图形处理单元(GPU)上加速序列搜索问题所需的内在并行计算，将我们的算法的性能和效率提高100到1000倍。最后，我们将设计专门的硬件架构来提高序列分析的速度，使其超过数量级，同时将其消耗的能量降低100倍或更多。我们的研究将广泛影响大规模基因组研究，如1000基因组计划、CAN-CER基因组图谱计划和ENCODE计划，不仅提高了他们快速得出结论的能力，而且还降低了他们与维护用于数据分析的计算簇相关的能源消耗和维护成本。如果我们的研究成功，可以消除序列分析对大型耗电计算集群/数据中心的依赖，从而使序列分析显著降低成本和能源效率，从而使序列分析能够由主流执行，而不需要构建大型计算基础设施。再加上测序技术的进一步进步，这一提议产生的研究可以帮助个人基因组学成为现实：药物基因组学的进步和应用将开启个性化医学时代。通过超快速、高能效和低成本的序列分析，这项研究可以通过分析目前无法用现有计算处理能力处理的太数据库序列数据，为无限数量的新发现铺平道路。 公共卫生相关性： 下一代测序(NGS)技术通过低成本和高通量的基因组测序承诺了预防性和个性化医学的时代。下一代测序的所有医学和遗传学应用的成功关键取决于计算技术的存在，这些技术可以快速、节能地处理和分析海量的序列数据，而不需要建立大型基础设施。这项提议的目标是通过结合增强的软件算法和专门的硬件加速器的好处来开发这种技术。