Leveraging Succinct Data Structures for DNA Sequence Mapping on FPGA

The sequencing capability of Next Generation Sequencing platforms continues to increase, but mapping short DNA fragments to large genomic sequences still constitutes a challenge in many bioinformatic applications. In this paper, we present BWaveR, a heterogeneous system leveraging FPGAs to increase the execution efficiency of the sequence matching process, based on succinct data structures that allow to encode large sequences in a minimal memory footprint, while still granting fast queries on them. Our design, evaluated on a Xilinx Alveo U200, results up to 4.9× faster and 26× more power-efficient than Bowtie2, a state-of-the-art equivalent software, run with 16-threads, without any loss in accuracy.