LoginSign Up
← Back to all sparks
M

MMseqs2

DEVOPS
Velocity0.0

Fast sequence search and clustering suite

github.com/soedinglab/MMseqs2 ↗

MMseqs2 put homology search on GPUs, then spent two releases making it behave

bioinformaticsgpu-accelerationsequence-searchhomologylicensing
◆Current state
Release 16 was the pivot: GPU-accelerated sensitive search on Turing-generation and newer CUDA hardware, shipped alongside a relicensing to MIT. The two releases since have been consolidation - Release 17 fixing GPU output corruption and a common prefilter crash, Release 18 restoring the custom substitution matrices that GPU support had cost users, making generated databases GPU-compatible, and adding a Forward-Backward aligner.
◆Where it's heading
The arc is a research tool absorbing a hardware shift. Each GPU release trades something away and buys it back later: Release 16 dropped custom substitution matrices, Release 18 restored them through a new lambda calculator. Underneath that, MMseqs2 keeps serving as the engine other tools are built on - Foldseek and ColabFold features appear in its release notes before they appear anywhere else.
◆Prediction
Expect GPU coverage to keep widening from search into the clustering and taxonomy workflows that still run on CPU, and the Forward-Backward aligner to gain the GPU path the rest of the alignment code now has. Further breaking database-format changes are likely as GPU compatibility propagates.

◆Recent moves

  1. 1y ago

    Custom substitution matrices restored; Forward-Backward aligner added

    Restores custom substitution matrices through a new lambda calculator, undoing the main regression from the GPU release, and makes databases GPU-compatible by default. A new Forward-Backward aligner and faster ARM64 alignment come with it.

    View source ↗
  2. 1y ago

    GPU output corruption and prefilter crash fixed

    Largely a stabilisation release for the GPU work: corrupted sequence output from the GPU version, a segmentation fault in easy-cluster, and a prefilter crash affecting many clustering runs are all fixed, and precompiled GPU binaries drop their glibc requirement to 2.17.

    View source ↗
  3. 1y ago

    MMseqs2 adds GPU-accelerated homology search

    ⚡ SPARK

    The release that moves MMseqs2's core competency onto the GPU, and relicenses the project to MIT at the same time. Everything in the two releases after it is consequence management.

    View source ↗
  4. 2y ago

    Ungapped prefilter mode and revised greedy clustering

    Adds an ungapped prefiltering mode that makes single-query searches efficient at low memory overhead, and revises the greedy clustering algorithm to pick representatives that respect the identity and coverage criteria. Solid pre-GPU throughput work.

    View source ↗
  5. 3y ago

    ColabFold and Foldseek features land; profile databases break

    A major release carrying features built for ColabFold and Foldseek plus a profile-profile preview, and breaking older profile databases, which must be regenerated. Shows MMseqs2 acting as shared infrastructure for the tools built on it.

    View source ↗
  6. 5y ago

    New taxonomy workflow for nucleotide-to-protein assignment

    ⚡ SPARK

    Introduces a taxonomy workflow that assigns taxonomic labels to nucleotide sequences by searching protein reference databases, optimised for contigs and considerably faster than what preceded it. A new capability rather than a faster version of an old one.

    View source ↗