Complete genome de novo assembly software for the emerging long read sequencing era

尽管短读段下一代测序（NGS）技术取得了巨大成功，但其固有的缺陷仍然存在。 无法建立长距离连接使得基本任务，如基因组闭合、单体型 定相和选择性剪接的转录物表征几乎是不可能的。两个长读段测序 提供商太平洋生物科学公司（PacBio）和牛津纳米孔技术公司（ONT）正在产生数据， 可以克服这些严重的缺点。PacBio能够产生10- 20 kb的读段， 特别是对封闭微生物基因组的采用增加，而且对欧洲基因组学也增加， 转录组学ONT的MinION设备是一种便携式实时测序平台，能够产生100 kb的 读取并已成功应用于微生物测序和病原体鉴定。ONT的 新型高通量仪器Promethion将于2016年发布，将有足够的产量用于 人类基因组规模实验。这两种技术的巨大潜力目前受到以下因素的阻碍： 高错误率（10-20%），这使得组装和共识调用的计算量非常大 挑战性已经开发了各种命令行软件程序来应对这些挑战， 它们通常需要大量的生物信息学专业知识和计算资源/知识， 二倍体基因组的关键障碍。随着长读段测序有望成为一个主要的 基因组学的资源，显然迫切需要集成易用的组装和分析 能够处理和利用这些数据的独特方面的软件。为此，我们制定了 原型从头汇编器基于我们的专利磁盘排序对齐（DSA）算法，可以汇编 将未校正的细菌基因组数据集合并为单个重叠群，在标准品上具有>99.2%的碱基准确度 台式电脑在不到3.5小时。组装器使用DSA确定的读段重叠来构建 装配字符串图，从该装配字符串图中，布局被馈送到一个新颖的共识生成器，该共识生成器被设计为最大化 从这个容易出错的数据的准确性。这一直接进入第二阶段提案的总体目标是将 原型为用于单倍体和二倍体基因组的完全可扩展的长读从头组装器。我们将首先 优化汇编程序组件的性能，为合并构建坚实的基础 识别两个姐妹之间的大的结构变异的基本二倍体感知能力 2）使共有碱基识别器适应于处理杂合SNV和小插入缺失，以及3） 利用数据的远程连接性来适当地对变体进行定相并产生准确的 单倍型序列最后，我们将利用这些工具来识别选择性剪接的转录本和等位基因- 从长阅读RNA-Seq数据的特异性表达。与DNASTAR 30年的交付历史相一致 易于使用的专家级软件，这个汇编程序将使任何用户访问这些革命性的长期阅读 测序技术以及即将到来的技术。

Despite the tremendous success of short read next-generation sequencing (NGS) technologies, their inherent inability to establish long range connectivity makes fundamental tasks such as genome closure, haplotype phasing and alternatively spliced transcript characterization all but impossible. Now, two long read sequencing providers, Pacific Biosciences (PacBio) and Oxford Nanopore Technologies (ONT), are producing data that can overcome these critical shortcomings. PacBio is capable of producing 10-20kb reads and has seen increased adoption for closing microbial genomes in particular, but also for eurkaryotic genomics and transcriptomics. ONT’s MinION device is a portable real-time sequencing platform capable of producing 100kb reads and has already been successfully applied to microbial sequencing and pathogen identification. ONT’s new high-throughput instrument, the PromethION, is being released in 2016 and will have sufficient output for human genome scale experiments. The tremendous potential of both technologies is currently hampered by high error rates (10-20%) which makes assembly and consensus calling extremely computationally challenging. Various command line software programs have been developed to tackle these challenges, but they typically require substantial bioinformatic expertise and computing resources/savvy and do not address the critical hurdles associated with diploid genomes. With long read sequencing poised to become a major resource for genomics, there is clearly an urgent need for integrated easy-to-use assembly and analysis software that can handle and exploit the unique aspects of this data. Toward that end, we have developed a prototype de novo assembler based on our patented Disk Sort Alignment (DSA) algorithm that can assemble an uncorrected bacterial genome data set into a single contig with >99.2% base accuracy on a standard desktop computer in less than 3.5 hours. The assembler uses DSA-determined read overlaps to construct an assembly string graph from which a layout is fed to a novel consensus generator designed to maximize accuracy from this error prone data. The overall goal of this direct to Phase II proposal is to transform the prototype into a fully scalable long read de novo assembler for both haploid and diploid genomes. We will first optimize the performance of the assembler components, building a solid foundation from which to incorporate the essential diploid-aware capabilities of 1) identifying large structural variation between two sister chromosomes, 2) adapting the consensus base caller to handle heterozygous SNVs and small indels and 3) exploiting the long range connectivity of the data to properly phase the variants and produce accurate haplotype sequences. Finally, we will leverage these tools to identify alternatively spliced transcripts and allele- specific expression from long read RNA-Seq data. Consistent with DNASTAR’s 30 year history of delivering easy-to-use expert level software, this assembler will give any user access to these revolutionary long read sequencing technologies and those to come.