TAD Detection (bioSkills)

A Claude Code skill for calling topologically associating domain boundaries from a Hi-C contact matrix, built around the diamond-window insulation score rather than a single hard domain partition.

   
Type Claude Skill
Supplier GPTomics bioSkills (community OSS, MIT)
Availability GA — part of the bioSkills collection
Pricing Free / OSS (MIT) — cooltools and HiCExplorer are installed separately (both open source)
Capabilities Read/Write — Claude runs the skill’s workflow locally (Python/Bash), not as an MCP tool
Verified works · 2026-08-13
Security cleared · 2026-08-13 — GPTomics/bioSkills MIT, no external credentials

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 "hi-c-analysis"

    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/hi-c-analysis/tad-detection ~/.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 Python Hi-C stack (the skill drives these; they are not bundled): 
    pip install "cooler>=0.10" "cooltools>=0.7" "bioframe>=0.7"
  • Optional — HiCExplorer 3.7+ for the hicFindTADs alternative: 
    conda install -c conda-forge -c bioconda hicexplorer

    Confirm with hicFindTADs --version. HiCExplorer pulls a large dependency tree; skip it if you only need the cooltools path.

What it does

Produces a continuous insulation track and a ranked boundary list, not just a BED file of domains:

  • Insulation score — cooltools.insulation() slides a diamond window along the diagonal of a balanced matrix and reports log2 insulation per bin, so a boundary is a valley whose depth is measurable.
  • Boundary strength — valley prominence, returned as boundary_strength_{W} alongside Li/Otsu-thresholded is_boundary_{W} flags, giving a ranking rather than a binary call.
  • Multi-scale window sweep — the skill sweeps a list of window sizes (roughly [3×bin, 5×bin, 10×bin, …]), from sub-TAD scale up to compartment-domain scale, because “TAD” is not a single-scale object.
  • Cross-condition comparison — compares the differential insulation score between conditions instead of intersecting two domain partitions.
  • Boundary annotation — supports CTCF-backed boundary annotation; overlap with other genomic features is routed to interval tooling, and domain rendering to the sibling Hi-C visualization skill.
  • HiCExplorer alternative — hicFindTADs as a second implementation.

Primary use cases: calling domain boundaries from a cooler, choosing an insulation window size, ranking and comparing boundaries across conditions.

Notes

Two input requirements will silently ruin a run if missed, and the skill states both. The cooler must be balanced (a stored weight column) — an unbalanced matrix returns all-NaN insulation, not an error. And a multi-resolution .mcool must be addressed with a single-resolution URI (file.mcool::/resolutions/10000), never the bare .mcool path. Balance first with cooler balance or cooler.balance_cooler().

The conceptual point the skill leads with is that the boundary is reproducible but the domain partition is not: different callers and different windows agree far better on where insulation dips than on how to segment the genome into domains, which is why the recommended output is a scored boundary track and why cross-condition work compares scores rather than partitions. Insulation is also treated as orthogonal to compartmentalization — a boundary call says nothing about A/B state.

Note the cooltools API shifted around 0.5→0.7 and standardized on view_df/viewframe arguments; pin cooltools>=0.7 or the documented call signatures will not match. Upstream skill front-matter name is bio-hi-c-analysis-tad-detection (tool_type: mixed, primary_tool: cooltools); upstream directory hi-c-analysis/tad-detection. Pairs with Chromatin Loop Calling (focal interactions rather than domain boundaries), A/B Compartment Analysis (the orthogonal megabase-scale layer), bedtools for boundary-feature overlap, and HOMER / JASPAR for CTCF motif orientation at boundaries.

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.