BitMapper2: A GPU-Accelerated All-Mapper Based on the Sparse q-Gram Index

Haoyu Cheng

doi:10.1109/TCBB.2018.2822687

BitMapper2: A GPU-Accelerated All-Mapper Based on the Sparse q-Gram Index

2019-05-01

10

PRE

AI

DOI：10.1109/TCBB.2018.2822687

Original

Original request for help

Save

Abstract

En 中文

The explosive growth of next-generation sequencing (NGS) read datasets drives a need for new faster read mappers. One class of read mappers, called all-mappers, is designed to identify all mapping locations of each read. Many all-mappers have been developed over the past few years, but they are either time-consuming or memory-consuming. Here, we present BitMapper2, a GPU-accelerated read mapper that reports all mapping locations of NGS reads. To make full use of the parallel processing capability of GPUs, BitMapper2 proposes the sparse q-gram index, which reduces the memory requirement and the data transfer time between GPU and CPU. We also design the filtration part and the verification part of BitMapper2 specifically for the architecture of GPU. In addition, BitMapper2 is still time-efficient and memory-efficient even if there is no GPU available. Experiments show that BitMapper2 was significantly faster than the state-of-the-art all-mappers, while requiring less space.

Keywords:

Read alignment

sparse q-gram index

edit distance

GPU

CUDA

next-generation sequencing

BitMapper2: A GPU-Accelerated All-Mapper Based on the Sparse q-Gram Index

Abstract

AI Summary

Journal

Organization