Jung Ho Ahn

De novo sequencing, a process to find the whole genome or the regions of a species without references, requires much higher computational power compared to mapped sequencing with references. The advent and continuous evolution of next-generation sequencing technologies further stress the demands of high-throughput processing of myriads of short DNA fragments. Recently announced sequence assemblers, such as Velvet, SOAPdenovo, and ABySS, all exploit parallelism to meet these computational demands since contemporary computer systems primarily rely on scaling the number of computing cores to improve performance. However, most of them are not tailored to exploit the full potential of these systems, leading to suboptimal performance. In this paper, we present ccTSA, a parallel sequence assembler that utilizes coverage to prune k-mers, find preferred edges, and resolve conflicts in preferred edges between k-mers. We minimize computation dependencies between threads to effectively parallelize k-mer processing. We also judiciously allocate and reuse memory space in order to lower memory usage and further improve sequencing speed. The results of ccTSA are compelling such that it runs several times faster than other assemblers while providing comparable quality values such as N50.