Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

bwa-mem2 (upstream)

bwa-mem2 is the direct predecessor of bwa-mem3 and the project from which the bwa-mem3 fork is derived. It was created at Intel’s Parallel Computing Lab by Vasimuddin Md and Sanchit Misra to accelerate the alignment algorithm originally written by Heng Li in bwa. bwa-mem2 achieves a 1.3–3.1x throughput improvement over the original bwa-mem by replacing key inner loops with vectorised implementations (SSE4.1, SSE4.2, AVX2, and AVX-512) and by switching to a more compact FM-index encoding. Its output is identical to bwa-mem at the alignment level, and it is distributed under the MIT license.

Lineage

The bwa alignment family has evolved through three generations, each building on the last:

  1. bwa — Written by Heng Li. Established the BWA-MEM algorithm, the SAM output format conventions, and the .bwt / .pac / .ann / .amb index layout.
  2. bwa-mem2 (Vasimuddin et al., Intel) — Replaced scalar inner loops with SIMD kernels; introduced the compact .bwt.2bit.64 and .0123 index formats; retained full output compatibility with bwa-mem.
  3. bwa-mem3 (Fulcrum Genomics fork) — Carries correctness fixes, performance improvements, new features (bisulfite alignment, mimalloc, ARM Neon), and expanded architecture support on top of the bwa-mem2 codebase. See What’s Different from bwa-mem2 for the full change catalog.

When you’d use it

Use bwa-mem2 directly when you need a stable, widely validated aligner with precompiled binaries available via Bioconda and the project’s GitHub releases page, and when you do not require the features or fixes that bwa-mem3 adds. bwa-mem2 is also the right choice when you are working in an environment where the bwa-mem3 fork has not yet been validated against your specific reference or sequencing library type.

How it relates to bwa-mem3

bwa-mem3 tracks bwa-mem2’s master branch and periodically rebases fork-carried commits on top of upstream changes. The What’s Different section documents every divergence between the two projects, and the Upstream PR status page tracks which bwa-mem3 changes have been proposed back to bwa-mem2. The goal is to keep the fork divergence minimal and to upstream as many fixes as practical.

  • GitHub: https://github.com/bwa-mem2/bwa-mem2
  • Citation: Vasimuddin Md, Sanchit Misra, Heng Li, Srinivas Aluru. “Efficient Architecture-Aware Acceleration of BWA-MEM for Multicore Systems.” IEEE IPDPS 2019.
  • License: MIT (with third-party components under their respective licenses)

See also: What’s Different from bwa-mem2 · Upstream PR status · bwa-mem3-bench · Citation