Hardening and Scaling Core Genomics Software

 描述(申请人提供)：下一代测序在生物学和疾病研究中普遍存在。越来越多的项目正在生成庞大的NGS数据集，越来越多的研究人员和学员正在使用复杂的软件来分析它们。我们提出了一项研究计划，目标是强化和改进Bowtie和Bowtie 2阅读比对软件工具。蝴蝶结1和2是由首席调查员创作和出版的。它们已成为广泛使用的关键工具。在给定测序数据集和参考基因组的情况下，读对齐器确定每个测序读数相对于参考源自何处。这将读数放在一个公共坐标系中，并允许进行下游分析，如变体调用和异构体组装。对齐在计算上具有挑战性，但它也是一种非常常见的需求。Bowtie 1和2用于分析RNA、亚硫酸氢盐、芯片、元基因组学和其他测序数据的软件工具的许多阶段。在目标1中，我们将支持和改进Bowtie 1和2。我们将(A)添加应用程序编程接口(API)，以便其他工具可以更轻松地访问Bowtie功能；(B)使Bowtie 1和2能够直接从公共档案获取数据；(C)扩展Bowtie 2，使其能够高效地处理长时间、易出错的读取，例如来自Nanopore测序仪的读取；以及(D)改进Bowtie 1和2，以更好地利用当前和即将到来的计算机体系结构上提供的许多处理器内核。在目标2中，我们解决了一个紧迫的可解释性问题：参照偏差。我们提出了一种短期和长期的解决方案；我们将(A)编译和传播“主等位基因”参考序列和创建新的参考序列的软件，(B)创建新的、高效的方法来即时编辑参考基因组和相关索引，(C)使Bowtie 1和2与CRGh38等图形形状的基因组兼容，以及(D)研究避免“爆炸”的基于最小程序的新的图形索引策略。在目标3中，我们将创建一个名为Rail的新软件系统，该系统能够扩展基于Bowtie的分析，以便研究人员和受训人员能够以容错、安全、可重现和廉价的方式分析非常大的公共数据集。Rail的设计是基于我们之前在可扩展的云计算工具Cross弓、Myrna和Rail-RNA上所做的工作。我们将与Galaxy项目的领导合作，整合Rail，以便Galaxy用户可以更轻松地构建具有高度可扩展组件的分析工具。目标2和目标3是通过科学合作来激发和验证的，这些合作解决了(A)等位基因特异性表达、(B)近交系和杂交品系的甲基化分析、(C)马赛克变异检测和(D)表达的重复元件的检测中的困难分析问题。Bowtie 1和2是开源软件；该项目生成的所有软件和数据都将在开源许可证下免费提供。

 DESCRIPTION (provided by applicant): Next-generation sequencing is ubiquitous in the study of biology and disease. More projects are generating vast NGS datasets and more investigators and trainees are using sophisticated software to analyze them. We propose a research program with the goal of hardening and improving the Bowtie and Bowtie 2 read alignment software tools. Bowtie 1&2 were created and published by the Principal Investigator. They have become widely used and crucial tools. Given a sequencing dataset and a reference genome, a read aligner determines where each sequencing read originated with respect to the reference. This puts reads in a common coordinate system and enables downstream analyses such as variant calling and isoform assembly. Alignment is computationally challenging, but it is also a very common need. Bowtie 1 and 2 are used at many stages in software tools for analyzing RNA, bisulfite, ChIP, metagenomics, and other sequencing data. In Aim 1 we will support and improve Bowtie 1&2. We will (a) add an application programming interface (API) so other tools can more easily access Bowtie functionality, (b) enable Bowtie 1&2 to obtain data directly from public archives, (c) extend Bowtie 2 to work efficiently with long, error-prone reads, such as those from Nanopore sequencers, and (d) improve Bowtie 1&2 to make better use of the many processor cores available on current and upcoming computer architectures. In Aim 2, we address a pressing interpretability issue: reference bias. We propose a mix of short- and long-term solutions; we will (a) compile and disseminate "major-allele" reference sequences, and software for creating new ones, (b) create novel, efficient methods for just-in-time editing of the referenc genome and associated index, (c) make Bowtie 1&2 compatible with graph- shaped genomes like CRGh38, and (d) investigate novel minimizer-based graph indexing strategy that avoids "blow-up." In Aim 3, we will create a new software system called Rail that enables scaling of Bowtie-based analyses so researchers and trainees can analyze very large public datasets in a manner that is fault tolerant, secure, reproducible, and inexpensive. Rail's design is based on our previous work on the scalable, cloud- enabled tools Crossbow, Myrna and Rail-RNA. We will work with leaders of the Galaxy project to incorporate Rail so that Galaxy users can more easily build analysis tools with highly scalable components. Aims 2 and 3 are motivated and validated by scientific collaborations addressing difficult analysis problems in (a) allele-specifi expression, (b) methylation analysis of inbred strains and crosses, (c) mosaic variant detection, and (d) detection of expressed repetitive elements. Bowtie 1&2 are open source software; all software and data generated by the project will be freely available under an open source license.