ParRefCom: Parallel Reference-based Compression of Paired-end Genomics Read Datasets
ParRefCom: Parallel Reference-based Compression of Paired-end Genomics Read Datasets
复制标题
DOI：
10.1145/3307339.3342171
复制
发表时间：
2019-09
期刊：
影响因子：
--
通讯作者：
Nagakishore Jammula;S. Aluru
中科院分区：
文献类型：
--
作者：
Nagakishore Jammula;S. Aluru
Transmission, storage, and archival of high-throughput sequencing (HTS) short-read datasets pose significant challenges due to the large size of such datasets. Constant improvements to HTS technology, in the form of increasing throughput and decreasing cost, and its increasing adoption amplify the problem. General-purpose compression algorithms have been widely adopted for representing read datasets in a compact form. However, they are unable to fully leverage the domain-specific properties of read datasets. In response, researchers proposed special-purpose compression algorithms which improve upon the compression efficiency of general-purpose compression algorithms. In this paper, we present ParRefCom, a parallel reference-based algorithm for compressing HTS genomics short-read datasets. HTS instruments are typically used to generate paired-end reads as they hold significance for biological analysis. In contrast to existing special-purpose compression algorithms, ParRefCom treats paired-end reads as first-class citizens. Owing to this treatment of paired-end reads, our algorithm is able to significantly improve compression efficiency over the state-of-the-art. More specifically, for a benchmark human dataset, the size of the compressed output is 21% smaller than that produced by the current best algorithm. Further, ParRefCom is scalable and its compression and decompression speeds are better than those of reference-free methods. Implementation : https://github.com/ParBLiSS/refcom
