Binder Discovery (ToolUniverse Claude Skill)

A ToolUniverse agent skill that runs a seven-phase small-molecule discovery workflow — from druggability assessment through docking and ADMET filtering — to hand back a ranked shortlist of candidate binders for a protein target.

   
Type Claude Skill (one of ToolUniverse’s pre-built agent skills)
Supplier Zitnik Lab, Harvard Medical School
Availability GA — part of the ToolUniverse skills collection (skills/tooluniverse-binder-discovery/)
Pricing Free / OSS (Apache-2.0); wraps public APIs (ChEMBL, BindingDB, PubChem, PDB, AlphaFold) plus NVIDIA NIM generative/docking endpoints
Capabilities Read-only — drives ToolUniverse tool calls; no data writes
Verified works · 2026-07-20
Security caution · 2026-07-20 — provenance/Apache-2.0 clear but docking/generation uses external NVIDIA NIM endpoints requiring a user NVIDIA_API_KEY

How to install

This skill calls ToolUniverse tools, so the ToolUniverse MCP server must be installed first (see the ToolUniverse page). Simplest registration:

claude mcp add --transport stdio tooluniverse -- uvx tooluniverse

Then add the skills:

  • Claude Code — install the whole skill collection (the skill resolves as tooluniverse-binder-discovery): 
    npx skills add mims-harvard/ToolUniverse
  • Manual / other agents — copy just this skill directory into your skills folder: 
    git clone https://github.com/mims-harvard/ToolUniverse
cp -r ToolUniverse/skills/tooluniverse-binder-discovery ~/.claude/skills/

    (replace ~/.claude/skills/ with your agent’s skills directory if you are not using Claude Code/Desktop.)

The skill sets disable-model-invocation: true upstream, so invoke it explicitly (e.g. ask Claude to “use the binder-discovery skill”) rather than relying on automatic dispatch.

What it does

Executes seven sequential phases to identify and prioritise drug-like compounds:

  1. Target validation — resolve IDs, assess druggability and binding sites (UniProt_search, MyGene_query_genes, OpenTargets_get_target_tractability_by_ensemblID, DGIdb_*).
  2. Known-ligand mining — extract bioactivity from curated databases (ChEMBL_get_target_activities, BindingDB_get_ligands_by_uniprot, GtoPdb_search_ligands, PubChem_search_assays_by_target_gene).
  3. Structure analysis — retrieve PDB/cryo-EM structures or predict them (PDB_search_similar_structures, get_binding_affinity_by_pdb_id, EMDB_search_structures, alphafold_get_prediction, InterPro_get_protein_domains).
  4. Docking validation — validate pocket geometry with a reference inhibitor (get_diffdock_info — NVIDIA NIM DiffDock; NvidiaNIM_boltz2).
  5. Compound expansion — similarity/substructure search and de novo generation (ChEMBL_search_similar_molecules, PubChem_search_compounds_by_similarity, NvidiaNIM_genmol scaffold hopping, NvidiaNIM_molmim analog generation).
  6. ADMET filtering — eliminate poor compounds on physicochemical/toxicity rules (ADMETAI_predict_physicochemical_properties, _predict_bioavailability, _predict_toxicity, _predict_CYP_interactions, ChEMBL_search_compound_structural_alerts).
  7. Docking, ranking, and report — score and prioritise the top ~20 candidates with literature-graded evidence.

Primary use cases: hit finding for a validated target, virtual-screening triage, generative analog design with ADMET gating.

Notes

It is a reasoning layer over ToolUniverse; without the MCP server registered, the tool calls fail. Several tools in phases 4–5 route to NVIDIA NIM generative/docking endpoints (DiffDock, Boltz-2, GenMol, MolMIM) — those calls need ToolUniverse’s NVIDIA NIM access configured, and may be rate-limited or unavailable without appropriate credentials; the ChEMBL/BindingDB/PubChem mining and ADMET-AI steps run against public APIs. ToolUniverse ships ~68 such skills; other drug-discovery workflows are catalogued separately.

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.