A New Solution For a (Scaff)Old Problem: an FPGA Approach

Broad access to genomic data is leading to improving the understanding, prevention, and treatment of human diseases as never before. In this context, genome assembly is a fundamental yet time-consuming step to enable DNA analysis on a large scale. This process involves various phases, and among these, scaffolding is one of the most computationally intensive. The exact matching process of sequences accounts for more than 86% of the total runtime of such a phase. In this work, we present a scalable, high-performance, and energy-efficient FPGA-based architecture tailored to the exact alignment step of the scaffolding process. Specifically, we propose an in-drop replacement for Bowtie, a widely employed literature tool for scaffolding, outperforming it by 2309x in speedup and 10443x in energy efficiency. Likewise, our best architecture achieves speedups and energy efficiency improvements ranging from 1.37x to 3364x and 5.05x to 7915x, respectively, when compared against hardware-based solutions in the State of the Art.