Drug Mechanism Research (ToolUniverse Claude Skill)

A ToolUniverse agent skill that answers “how does this drug actually work?” — chaining primary target, off-target binding, pathway context, the FDA label’s own mechanism language, and pharmacogenomic modifiers into one evidence-tiered report.


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-drug-mechanism-research/`)
Pricing	Free / OSS (Apache-2.0); wraps public sources (Open Targets, ChEMBL, KEGG, Reactome, WikiPathways, STRING, DailyMed, PharmGKB, PubMed, Europe PMC)
Capabilities	Read-only — database and literature queries; runs no computation on your data
Verified	works · 2026-08-13
Security	cleared · 2026-08-13 — mims-harvard/ToolUniverse Apache-2.0, read-only public sources

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-drug-mechanism-research):
```
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-drug-mechanism-research ~/.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 drug-mechanism-research skill”) rather than relying on automatic dispatch.

What it does

Runs an eight-step investigation:

Drug resolution — name to canonical IDs (ChEMBL ID via OpenTargets_get_drug_id_description_by_name, PharmGKB ID via PharmGKB_search_drugs) before any downstream query.
Primary target — the bound protein and the action type (inhibition, agonism, antagonism) via OpenTargets_get_drug_mechanisms_of_action_by_chemblId and ChEMBL_get_drug_mechanisms.
Off-target assessment — additional proteins bound at clinical concentrations, via ChEMBL_get_target_activities and STRING_get_interaction_partners, as the explanation for side effects and interactions.
Pathway mapping — the target’s position in KEGG / Reactome / WikiPathways, distinguishing an upstream target (broad downstream effects) from a downstream one (narrow effects).
Regulatory view — the FDA label’s own mechanism narrative parsed out of DailyMed, kept alongside the molecular detail rather than replaced by it.
Pharmacogenomics — variants that modify response, focused on CPIC Level A/B gene–drug pairs and FDA-labeled biomarkers.
Literature evidence — PubMed and Europe PMC support for each claimed mechanism.
Integration — a structured report ordered by an explicit evidence hierarchy.

Rules the skill applies: nanomolar binding affinity marks a primary target while micromolar affinity marks an off-target effect at clinical dose; when several targets converge on one pathway, that pathway is reported as the true mechanism, and targets in genuinely different pathways are reported separately rather than merged.

Findings are graded on a four-tier hierarchy — T1 FDA labels, CPIC Level A guidelines, FDA PGx biomarkers; T2 ChEMBL mechanisms with literature references and binding-affinity data; T3 database MOA entries and pathway databases; T4 PubMed / Europe PMC articles.

Primary use cases: explaining how an approved drug works, rationalizing an unexpected side effect as off-target binding, mechanism-based combination design, drafting the mechanism section of a report.

Notes

It is a reasoning layer over ToolUniverse; without the MCP server registered, the tool calls fail.

The tier system is doing real work here — a mechanism supported only by a T3 database MOA entry is a database assertion, not a measurement, and the skill is designed to say so rather than flatten everything into one narrative. Binding-affinity calls depend on ChEMBL’s assay records, which mix assay types and conditions; treat a nanomolar-vs-micromolar split as a triage heuristic, not a measured selectivity ratio.

Complements Drug Research (broader profile of one drug), Drug-Target Validation (evidence that a target is worth pursuing), and Drug-Drug Interaction (the interaction surface this skill’s off-target step feeds). The individual sources are catalogued separately: Open Targets, ChEMBL, KEGG, Reactome, STRING, DailyMed, ClinPGx / PharmGKB. ToolUniverse ships ~68 such skills; other 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.