Hardware-accelerating the BLASTN bioinformatics algorithm using high level synthesis

Abstract

BLASTN is one of the most known algorithms used for biological sequence analysis in bioinformatics. This algorithm is highly optimized for similarity searches. However, the significance increase in the size of genomic databases causes performance degradation in the search algorithm. Thus, there is an increasing need to accelerate this algorithm. This paper introduces a new hardware approach to accelerate BLASTN using high level synthesis. Our approach takes advantage of the high level synthesis capabilities in designing complex systems by simply writing the algorithm functionalities in high level language. Experimental results show a speedup up to 100x over software. Moreover, we proved the feasibility of our proposed HLS implementation through a comparison with similar algorithms. Our HLS BLASTN achieves an average speedup of 70x over the NCBI BLASTN implementation and a speedup of 11x over the Mercury BLASTN implementation. We conclude that high level synthesis is an appealing approach to accelerate the BLASTN algorithm to satisfy the high performance need of biological searches.