Accelerating genomic analysis for time critical clinical applications

总结/摘要 基因组规模的DNA测序彻底改变了精准医学的实践， 成本今天，在大约一天内对整个人类基因组进行测序是可能的;然而，生物信息学 分析通常需要数天或数周时间，已成为成功利用 在时间关键型应用中的基因组测序，例如，用于识别患者的基因组脆弱性， 在临床相关的时间范围内合理的癌症治疗选择。这个项目的首要目标是 一项提案是通过异构计算大大加快基因组分析算法 技术.在这里，我们将集中于基因组分析的一个关键方面，即变异识别，并设置 我们的雄心勃勃的目标是在不到1000年的时间内完成对60 X覆盖率的Illumina全基因组测序数据集的分析。 10分钟，远远快于目前的技术水平。虽然这里只适用于一个分析任务， 实现如此高的加速将证明我们正在开发的技术 该建议也可推广到许多其它基因组分析任务。我们的方法是首先 加速最广泛可重用的软件组件，以最大限度地提高基因组分析工具的价值 开发人员社区，他们将能够将这些组件集成到自己的工具中。 有了这些可重用的软件组件，我们将加速FreeBayes变量调用工具。FreeBayes是一个 广泛使用的生殖系变异和体细胞突变检测工具，因此加速将有利于大量的 用户受众。该软件是在我们自己的实验室开发的，因此我们非常熟悉 它的算法和代码库，使我们能够在这个探索性项目中取得成功。如果成功，我们的技术 将适用于加速许多目前耗时的分析任务。因此，分析师将 能够在几分钟内完成复杂的数据处理任务，作为其交互式分析会话的一部分 而不是一个批处理的后台进程，并在此之后立即完成手动结果审查; 完整的分析过程足够快，适用于时间紧迫的临床应用。

SUMMARY / ABSTRACT Genome-scale DNA sequencing has revolutionized the practice of precision medicine, at dramatically reduced cost. It is possible today to sequence an entire human genome in roughly one day; however, bioinformatic analysis typically takes days or weeks, and has emerged as the major bottleneck for successfully utilizing genome sequencing in time-critical applications, e.g. for identifying the genomic vulnerabilities of a patient’s tumor for rational cancer treatment selection within a clinically relevant timeframe. The overarching goal of this proposal is to dramatically speed up genomic analysis algorithms via heterogeneous computing techniques. Here we will focus on one critical aspect of genomic analysis, i.e. variant calling, and set the ambitious goal of completing the analysis of a 60X-coverage Illumina whole genome sequencing dataset in under 10 minutes, far faster than the current state of the art. Although here applied to only one analysis task, accomplishing such a high degree of acceleration would demonstrate that the techniques we are developing in this proposal are also generalizable across many other genomic analysis tasks. Our approach is to first accelerate the most widely reusable software components, to maximize value for the genomic analysis tool developer community, who will then be able to integrate these components into their own tools. With these reusable software components, we will accelerate the FreeBayes variant caller tool. FreeBayes is a widely used germline variant and somatic mutation detection tool, and therefore acceleration will benefit a large user audience. This software was developed in our own laboratory, and therefore we are intimately familiar with its algorithms and code base, positioning us for success in this exploratory project. If successful, our technique will be applicable for accelerating many, currently time-consuming analysis tasks. As a result, analysts will be able to finish sophisticated data processing tasks within minutes, as part of their interactive analysis session rather than a batched background process, and complete manual result review immediately after; rendering the complete analysis process sufficiently fast for time-critical clinical applications.