Published July 27, 2025 | Version 18-8cc5c
Software Open

soedinglab/MMseqs2: MMseqs2 Release 18-8cc5c

Authors/Creators

  • 1. Seoul National University
  • 2. LJK-GINP
  • 3. ELKMO
  • 4. Max Planck Institute
  • 5. Max-Planck institute for biophysical chemistry
  • 6. @common-workflow-language
  • 7. @nvidia
  • 8. @NeelyxLabs
  • 9. Sunagawa Lab @ ETH Zürich
  • 10. University of São Paulo
  • 11. Southern University of Science and Technology
  • 12. Zhejiang University

Description

MMseqs2 Release 18 at a glance: new Forward–Backward aligner, re-enabled substitution matrix parameter estimation, faster ARM64 alignments, improved GPU support.

Breaking changes

  • CMake ≥ 3.15 is now required (da0b2c3f).
  • gpuserver no longer accepts the --gpu parameter (3b5d13e6).

New features and enhancements

  • New Forward‑Backward (FWBW) aligner (003fabcd, fb687b7b) by @Gyuuul2 @elpis51613 @lasseReifenrath
  • Custom substitution matrices are supported again through a new lambda calculator (5ebd6e9d, efad625b, b76ebc43) by @edawson.
  • Proximity‑aware pairing: pairaln can now match sequences that are physically close in accession space (--proximity, pairaln list; e019185b, c9107bac, 835acb97, 1970db60, f6e9636d, 60a894b4).
  • createdb and databases accept --gpu parameter to directly produce GPU databases (05789398, 90ee5423).
  • Speedup aarch64 SIMD alignment with new/improved simd_any, simd_eq_all, simd_hmax* instructions (103fe796).
  • aarch64 GPU binaries mmseqs-linux-gpu-arm64.tar.gz are now built with Clang 20 and wide Neon registers for additional speedup (62cf4d09, 9564601c, 1668032c). mmseqs-linux-arm64.tar.gzis still the old slower version.
  • createdmptaxonomy allows converting taxonomy databases back to .dmp files (bc0f9cb8).
  • taxonomyreport can now emit one database per query with --report-mode 3 (8284a8b1, 033d5f5f).
  • Reduce thread start overhead in expandaln, pairaln, subtractdbs and unpackdb (18d8ddc2).

Bug fixes

  • Improve error handling for createindex with only ungapped prefilter (9668e96c, 829003a8).
  • Fix precomputed index being slightly too large (b98f2076).
  • expandaln now skips entries lacking alignments (9c132756) and finds correct representatives (#691, 8783404e).
  • clusterupdate preserves members correctly (#961, e7f58527, 296d9122).
  • MPI nucleotide clustering was crashing (defe1afc)
  • SAM start coordinate was wrong sometimes (c13eef01)
  • Systems with old mawk could result in corrupted databases due to large‑int printing (eaecacf4)
  • Wrong createdb mode message (#955, 48143e74)
  • Fix createdb could crash if FD 0 was closed (99a025e6). Thanks @jnooree

Developer notes

  • GPU binaries are now compiled with prebuilt GPU containers available at https://github.com/steineggerlab/build-containers (and other CI improvements; 12a0d9ac, 887b4411, 4289ae3e, 0eb08bf6).
  • GPU regression tests added to CI (bebc9563, 4289ae3e).

Files

soedinglab/MMseqs2-18-8cc5c.zip

Files (14.4 MB)

Name Size Download all
md5:d8acf62fc7258a924c994492b6791f1d
14.4 MB Preview Download

Additional details

Related works

Is supplement to
Software: https://github.com/soedinglab/MMseqs2/tree/18-8cc5c (URL)

Software

Repository URL
https://github.com/soedinglab/MMseqs2