Multiple Sequence Alignment (bioSkills)

A Claude Code skill that picks the alignment algorithm from the two facts that determine it — how many sequences and how diverged — instead of running the same default on every dataset.

   
Type Claude Skill
Supplier GPTomics bioSkills (community OSS, MIT)
Availability GA — part of the bioSkills collection
Pricing Free / OSS (MIT). The aligners install separately and are all free/OSS: MAFFT (BSD), MUSCLE5 (GPL-3.0), Clustal Omega (GPL-2.0), T-Coffee (GPL-3.0)
Capabilities Read/Write — Claude runs the skill’s workflow locally (Bash/Python), not as an MCP tool
Verified works · 2026-08-17
Security caution · 2026-08-17 — GPTomics/bioSkills is now archived upstream; MIT and skill dir confirmed unchanged

How to install

bioSkills is not an npm package — skills are plain markdown/code read directly by the agent. Clone the repo, then either run the installer for the whole category or copy the single skill directory.

  • Claude Code — clone and install via the bundled script: 
    git clone https://github.com/GPTomics/bioSkills
cd bioSkills
./install-claude.sh --categories "alignment"

    The installer copies matching skills into ~/.claude/skills/ (default target). Use ./install-claude.sh --list to preview the skills first.

  • Claude Code / other agents — copy just this one skill: 
    cp -r bioSkills/alignment/multiple-alignment ~/.claude/skills/

    (run from inside your clone — the previous step left you in bioSkills/; otherwise replace bioSkills/ with the absolute path of your clone, e.g. /Users/you/repos/bioSkills).

  • Prerequisites — the aligners are CLI tools the skill drives, none are bundled. All are on bioconda (versions checked 2026-08-15, each satisfying the skill’s stated minimum): 
    conda install -c conda-forge -c bioconda mafft muscle clustalo t-coffee

    mafft 7.525 (skill wants 7.520+), muscle 5.3 (5.1+), clustalo 1.2.4 (1.2.4+), t-coffee 13.46.0 (13+). Confirm with mafft --version, muscle -version, clustalo --version, t_coffee -version.

  • Codon-alignment extras, only if you are heading for dN/dS: 
    conda install -c conda-forge -c bioconda pal2nal prank macse

    (Unverified — bioconda versions for these three were not checked on this pass; run conda search before relying on them.)

  • Python side — BioPython 1.83+: 
    pip install "biopython>=1.83"

    Note the API change the skill flags: Bio.Align.Applications was removed in BioPython 1.86, so wrapper classes no longer exist and the aligners must be invoked with subprocess directly.

What it does

  • Algorithm selection by size — reproduces MAFFT’s --auto decision boundaries explicitly so you can see which branch you are on: < 200 sequences → L-INS-i (local pairwise iterative, the accuracy option); 200–500 → FFT-NS-i truncated to --maxiterate 2; 500–2,000 → FFT-NS-2 progressive; 2,000–50,000 → single-pass FFT-NS-2; > 50,000 → PartTree.
  • Tool capacity — MAFFT L-INS-i tops out around 200 sequences at highest accuracy; MUSCLE5 PPP (-align) handles about 1,000 at highest accuracy; MUSCLE5 -super5 scales past 100,000; Clustal Omega has a published benchmark at 190,000.
  • Confidence assessment — per-column reliability with GUIDANCE2 or T-Coffee TCS before anything downstream consumes the alignment, then masking rather than silent acceptance.
  • Codon-aware alignment — protein-first with MAFFT then threading nucleotides back with PAL2NAL for clean orthologs; PRANK +F codon for indel-rich paralogs; MACSE v2 where frameshifts or pseudogenes are present; HyPhy’s pre-msa.bf / post-msa.bf for BUSTED/MEME-grade input.
  • Six-stage workflow — homology check (BLAST E-value < 1e-5, drop non-homologous sequences) → algorithm selection → alignment → confidence scoring → low-confidence masking → downstream analysis.

Primary use cases: building a defensible alignment for phylogenetics or selection analysis, choosing between MAFFT modes on a large family, preparing codon alignments for dN/dS.

Notes

The load-bearing threshold is the twilight zone, stated as a table: above 40% protein identity any MSA tool will do; 25–40% requires iterative methods with GUIDANCE2 validation; 20–25% calls for profile–profile methods; and below 15–20% the sequence signal is dominated by noise and the correct move is to switch to structural alignment altogether. That last row is the one most often ignored — a low-identity alignment is not merely imprecise, it is a different alignment on every rerun.

The sharpest warning concerns codon work: standard nucleotide MSA tools break reading frames and produce systematically incorrect dN/dS estimates and false-positive selection signals even on clean simulated data (Fletcher & Yang 2010). If the destination is PAML or HyPhy, a nucleotide-level MAFFT run is not an acceptable input, which is why the skill routes through protein alignment plus PAL2NAL, PRANK codon mode, or MACSE.

Upstream skill front-matter name is bio-alignment-multiple; upstream directory alignment/multiple-alignment; primary_tool is MAFFT. This is the first step of a four-part chain: build here, then Alignment Trimming, then MSA Statistics, with Structural Alignment taking over below the identity floor. Trees are built in Phylogenetics (which bundles its own MAFFT + IQ-TREE 2 path for the simple case), covariation testing of RNA alignments in Covariation Analysis, and profile-HMM/covariance-model search in ncRNA Search and BLAST.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.