Hardware Acceleration of BWA-MEM Genome Mapping Application

Abstract

Next Generation Sequencing technologies have had a tremendous impact on our understanding of DNA and its role in living organisms. The cost of DNA sequencing has decreased drastically over the past decade, leading the way for personalised genomics. Sequencers provide millions of fragments of DNA (termed short reads), which have to be aligned aginst a reference genome to reconstruct the full information of the DNA molecule under study. Processing short reads requires large computational resources, with many specialised computing platforms now being used to accelerate software aligners. We take up the challenge of accelerating a well know sequence alignment tool called Burrows Wheeler aligner on the Convey Hybrid Computing platform, which has FPGAs as co-processors. The focus of the research is to accelerate the BWA-MEM algorithm of the Burrows Wheeler aligner on the Convey HC-2 platform. The implementation was carried out using the Vivado HLS tool. The architectures proposed are targeted to overcome the memory bottleneck of the application. Two architectures are proposed, the Base architecture and the Batch architecture meant to address the memory bottle- neck. The architectures provide possibilities of further pipelining and implementation of more cores, which is expected to provide better performance than the current implementation.