Power efficient special processor design for burrows-wheeler-transform-based short read sequence alignment

摘要

Because of the advancement of Next Generation Sequencing technology, DNA short reads can be generated in parallel at very high-throughput rates. One of the new challenges in bioinformatics is how to map these high volume short reads to the target reference accurately and efficiently. Besides the conventional software approaches, using a special processor for sequence alignment is a promising alternative. In this work, we present a Burrows-Wheeler-Transform-based hardware architecture and implement it on a special processor using 90 nm TSMC technology. With techniques including Forward String Sorting, Depth-First Search, and Extension Modules, our hardware is very efficient in terms of time and power. When compared to the hundred-Watt-level software approach, the proposed solution needs only less than 600 mW to achieve the same speed performance.