High Accuracy Short Reads Alignment Using Multiple Hash Index Tables on FPGA Platform

Next-generation sequencing can preliminarily solve many problems in the biological area and it plays a more and more important role in bioinformatics. An important step of next-generation sequencing is mapping the query reads to a long reference DNA sequence. Nevertheless, many existing read alignment tools do not have high mapping accuracy. BWA-MEM is one of the most popular tools used for mapping short reads to reference sequences and it can be considered a good read alignment tool with high accuracy. In this paper, we focus on designing a high accuracy short reads alignment system utilizing the read mapping algorithm based upon seed-and-extend type hash tables. To replace the BWA-MEM, we present a heterogeneous computing platform to construct our hash based mapping system and put forward new algorithms to replace corresponding BWA-MEM algorithms. Our system consists of a host processor to execute logical operations, and a coprocessor FPGA to accelerate computation speed. In addition, we experiment on our proposed method and compare the precision with related BWA-MEM methods. The experiment results show that we achieve more than 99 percent similarity of the intermediate results and 99.90 percent similarity of the final results with BWA-MEM. Finally, the comparison demonstrates that both BWA-MEM and our hash based mapping system can get the same output without the random selection.