Drugmechdb

Skill BioTender-max/awesome-bio-agent-skills/skills/drugclaw/drugmechdb

A curated collection of AI agent skills for biomedical research, covering genomics, proteomics, single-cell analysis, clinical AI, and protein design.

Install

npx -y skills add BioTender-max/awesome-bio-agent-skills --skill drugmechdb

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.

What its author says it does

Copied from the file, not written here

Query the DrugMechDB drug mechanism-of-action database. Use whenever the user asks about drug mechanisms, drug-to-disease paths, biological targets of a drug, or wants to look up any biomedical entity (drug name, protein, disease, DrugBank ID, MESH ID, UniProt ID, GO term, etc.) in DrugMechDB.

SKILL.md

2.9 KB, as published. Nobody here has run it

DrugMechDB Query Skill

Search drug mechanism-of-action paths by entity name or ID. Each path is a directed graph: Drug → (intermediates) → Disease, with typed nodes and labeled edges.

Data

Source: indication_paths.json
Path: resources_metadata/drug_mechanism/DRUGMECHDB/indication_paths.json
Records: ~4846 mechanism paths, ~32k relationships

Entity auto-detection

Input pattern	Detected as	Example
`DB:DB00619`	DrugBank ID	exact on node/graph IDs
`MESH:D015464`	MESH ID	exact on node/graph IDs
`UniProt:P00519`	UniProt protein	exact on node IDs
`GO:0006915`	GO term	exact on node IDs
`CHEBI:`, `HP:`, `UBERON:`, `CL:`, `reactome:`, `InterPro:`, `PR:`, `taxonomy:`	respective types	exact on node IDs
anything else	free text	substring match on drug/disease/node names

API

Function	Signature	Returns
`load(path)`	path to JSON	`list[dict]` — full database
`build_index(db)`	loaded db	`(by_id, by_name, by_drug, by_disease)` dicts for O(1) lookup
`search(db, entity, index=None)`	single query string	`list[dict]` — matching paths
`search_batch(db, entities, index=None)`	list of query strings	`dict[str, list[dict]]`
`summarize(paths, entity)`	search results	compact multi-line text
`to_json(paths)`	search results	list of flat dicts (id, drug, disease, nodes, links)

Node types (14)

BiologicalProcess, Cell, CellularComponent, ChemicalSubstance, Disease, Drug, GeneFamily, GrossAnatomicalStructure, MacromolecularComplex, MolecularActivity, OrganismTaxon, Pathway, PhenotypicFeature, Protein

Quick usage

import drugmechdb_query as dq

db = dq.load()  # uses default DATA_PATH
idx = dq.build_index(db)  # optional, recommended for repeated queries

# Single query — by name or ID
paths = dq.search(db, "imatinib", idx)
paths = dq.search(db, "UniProt:P00519", idx)
paths = dq.search(db, "MESH:D003920", idx)
print(dq.summarize(paths, "imatinib"))

# Batch query
results = dq.search_batch(db, ["metformin", "MESH:D003920", "asthma"], idx)
for entity, paths in results.items():
    print(dq.summarize(paths, entity))

# JSON export
print(dq.to_json(paths))

Output structure per path

graph:  { _id, drug, disease, drugbank, drug_mesh, disease_mesh }
nodes:  [{ id, label, name }, ...]
links:  [{ source, target, key }, ...]

key examples: decreases activity of, causes, positively regulates, treats, increases expression of, etc. (66 relation types total).