BindingDB Query Skill

SkillDatabases & data

Query the BindingDB drug-target binding affinity database. Use whenever the user asks about protein-ligand binding data, affinity measurements (Ki, IC50, Kd, EC50), or wants to look up binding partners for a UniProt ID, PDB ID, or compound SMILES string.

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Account requirements not reviewed. Check the skill instructions before use; ahel provides instructions and does not run this skill.

Add ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

Then ask your AI: use the BindingDB Query Skill skill

What this skill tells your AI

The instructions your AI receives, as published by qsong-github/drugclaw in skills/dti/bindingdb/SKILL.md and read by ahel’s review.

Search BindingDB binding affinity records by any entity. Auto-detects type by pattern:

Input Pattern	Detected As	Example	API Endpoint
`P35355`, `Q9Y233`	UniProt ID	`P00533` (EGFR)	`getLigandsByUniprots`
`1Q0L`, `3ANM`	PDB ID (4-char, digit-leading)	`1Q0L`	`getLigandsByPDBs`
contains `=()#[]@/\`	SMILES string	`CC(=O)Oc1ccccc1C(O)=O`	`getTargetByCompound`
fallback	treated as UniProt	—	`getLigandsByUniprots`

API

Function	Input	Returns
`search(entity, cutoff)`	single entity string	dict with `entity`, `type`, `hits`, `affinities`
`search_batch(entities, cutoff)`	list of entity strings	dict[str, search_result]
`summarize(result)`	search() output	compact multi-line text
`to_json(result)`	search() output	list[dict] of affinity records
`query_by_uniprot(ids, cutoff)`	UniProt ID(s), nM cutoff	list[dict]
`query_by_pdb(ids, cutoff, identity)`	PDB ID(s), nM cutoff, % identity	list[dict]
`query_by_smiles(smiles, cutoff)`	SMILES, similarity 0–1	list[dict]

Parameters

cutoff (int): affinity threshold in nM (default 10 000). Entries with IC50/Ki/Kd ≤ cutoff are returned.
identity (int, PDB only): sequence-identity cutoff in percent (default 92).
Results are capped at 50 per query for LLM readability.

Usage

See if __name__ == "__main__" block in 26_BindingDB.py for runnable examples covering: UniProt single query, PDB query, SMILES compound query, batch query, and JSON output.

Key Fields in Each Affinity Record

Field	Description
`query`	Target protein name
`monomerid`	BindingDB compound ID
`smile`	SMILES structure of ligand
`affinity_type`	Ki, IC50, Kd, or EC50
`affinity`	Value in nM
`pmid`	PubMed ID of source
`doi`	DOI of source publication

Data Source

Database: BindingDB (https://www.bindingdb.org/)
Size: 3.2M data points, 1.4M compounds, 11.4K targets
Access: Public REST API (JSON), no authentication required
Citation: BindingDB in 2024: a FAIR knowledgebase of protein-small molecule binding data. Nucleic Acids Research, 53(D1), D1633 (2025). DOI: 10.1093/nar/gkae1199

Signals

GitHub stars: 116
Forks: 3
Last commit: Aug 2026

Advanced

Item type: skill
Key: bindingdb-query
Source: github.com/qsong-github/drugclaw

github.com/qsong-github/drugclaw