更高效的PacBio长read纠错算法的研究

The long reads from the third generation PacBio sequencing technology have been more and more widely used in various sequencing projects, and an important computational step is reducing the about 15% error rate of the reads. The current three types of error correction algorithms based on short reads, based on short read assemblies, and with long reads alone can reduce the error rate to 1%, but still have some problems. (1) To resolve the problem that all the current algorithms have relatively low rate of maintained bases after correction, we propose to study a high base maintenance rate algorithm based on contigs assembled from short reads. The algorithm will precisely find the alignments between contigs assembled from short reads and the long reads, by constructing a weighted directed graph and defining and solving a combinatorial optimization problem. (2) To resolve the problem that the current algorithms based on short read assemblies do not have the correction function targeting chimeric errors, we propose to study a chimeric error correction algorithm also based on contigs assembled from short reads. The algorithm will distinguish between chimeric errors from the long reads and the contigs, by comparing the long and short read supports upon the long reads and the contigs, respectively. (3) To resolve the problem that the current algorithm with long reads alone requires very high long read coverage, we propose to study a moderate coverage requirement algorithm. The algorithm will align repetitive regions to each other with reasonable runtime, by indexing and aligning k-mers from the long reads.

第三代PacBio测序技术的长read已越来越广泛的应用于各类测序项目中，而降低其约15%的错误率是必要的计算步骤。当前各类纠错算法可把长read的错误率降至1%，但也存在着一些问题。（1）针对各类算法碱基保留率较低的问题，我们提出研究基于短read contig的高碱基保留率算法，拟通过构建加权有向图和定义解决组合优化问题，来精确找到contig到长read的正确比对结果。（2）针对基于短read拼接的算法缺少chimeric错误纠错功能的问题，我们提出研究基于短read contig的chimeric错误纠错算法，拟通过比较长短read对长read和contig的支持度，来区分两者的chimeric错误。（3）针对自纠错算法对长read覆盖度要求太高的问题，我们提出研究覆盖度要求适中的长read自纠错算法，拟通过索引和比对长read的k-mer，来保证大量重复区域比对的运行时间。

PacBio公司的第三代单分子实时测序技术已经被越来越广泛的应用于各类基因组测序项目之中，但是这一技术生成的长读长有15%左右的错误率，且包含导致错误拼接的嵌合错误。我们主要研究PacBio长读长的纠错算法，取得了以下三方面的研究成果。（1）长读长的高通量纠错算法HALC：该算法借助于与长读长来自同一物种的短读长和其拼接所得的重叠群对长读长纠错，取得了比其它纠错算法高6.7-41.1%的通量。（2）基于参考基因组和长短读长的重叠群错误拼接检测算法ReMILO：该算法综合使用来自相似物种的参考基因组和长读长来检测和纠正重叠群的错误拼接，以及长读长的嵌合错误，可以比其它算法多检测出11.6-98.5%的错误拼接。（3）长读长的快速、高通量自纠错算法HALS：该算法对当前最快速的长读长自纠错算法MECAT进行改进，取得了比MECAT高28.1-230.2%的通量。我们共发表3篇高水平论文，包括以项目负责人为第一作者和通讯作者的1区SCI期刊论文一篇和2区SCI期刊论文一篇，以及以项目负责人为第一作者的顶级会议子会论文一篇。

内容获取失败，请点击重试