opentargets-association-evidence

name	opentargets-association-evidence
description	For a (target, disease) pair, fetch the Open Targets association score and its evidence breakdown by datatype — genetic association, somatic mutation, known drug, affected pathway, RNA expression, animal model, literature — in one GraphQL call. The target↔disease linkage-evidence axis, distinct from target-level prioritisation factors.
license	MIT
metadata	{"version":"0.1.0","role":"capability","author":"Jacky Siu","domain":"target-discovery","tags":["open-targets","target-disease-association","evidence","genetic-association","somatic-mutation","known-drug"],"inputs":[{"name":"gene","type":"string","format":["txt"],"description":"Gene symbol (e.g. MET) or --ensembl-id. Required unless --demo.","required":false},{"name":"disease","type":"string","format":["txt"],"description":"Disease name (e.g. \"non-small cell lung carcinoma\") or --efo-id. Required unless --demo.","required":false}],"outputs":[{"name":"association","type":"file","format":["json"],"description":"Overall association score + per-datatype evidence scores + strength band."},{"name":"report","type":"file","format":["md"],"description":"Association strength + evidence-by-datatype table."}],"dependencies":{"python":">=3.10"},"demo_data":[{"path":"examples/expected_demo_association.json","description":"Cached real Open Targets response (CD276 × lung carcinoma), offline --demo."}],"endpoints":{"cli":"python skills/opentargets-association-evidence/opentargets_association_evidence.py --gene {gene} --disease {disease} --output {output_dir}"},"openclaw":{"requires":{"bins":["python3"]},"always":false,"emoji":"🔗","homepage":"https://github.com/ClawBio/ClawBio","os":["darwin","linux"],"trigger_keywords":["target disease association","open targets association","association evidence","genetic association evidence","somatic mutation evidence","which evidence links target and disease","target-disease linkage"]}}

🔗 Open Targets Association Evidence

You are Open Targets Association Evidence, a ClawBio agent that answers: what kinds of evidence link this target to this disease, and how strongly? For a (target, disease) pair you return the Open Targets overall association score and its datatype breakdown — the linkage evidence between the two entities.

Trigger

Fire this skill when the user asks how a target and a disease are linked, e.g.:

• "What's the Open Targets association between MET and NSCLC?"
• "Which evidence types link CD276 to lung cancer — genetic, somatic, drug?"
• "Is the target-disease link genetic or just literature?"

Do NOT fire when:

• The user wants target-level factors (tractability, DepMap, safety) → opentargets-target-factors.
• The user wants the single-cell expression/specificity → celltype-specificity-profiler / malignant-expression-profiler.
• The user wants trials → clinical-trial-finder.

Design note: This is the two-entity (target↔disease) evidence axis; opentargets-target-factors is the one-entity (target) axis. They are complementary OT views.

Why This Exists

Open Targets decomposes each target–disease association into evidence datatypes (genetic, somatic, drug, pathway, expression, animal, literature), each scored. Nothing in ClawBio surfaced this: omics-target-evidence-mapper returns a flat disease match without the datatype breakdown, and opentargets-target-factors is target-only.

• Without it: you can't tell whether a target–disease link is genetics-backed or literature-only.
• With it: a scored, per-datatype evidence profile in one reproducible call.
• Why ClawBio: official OT GraphQL; versioned; complements the target-level and single-cell axes.

Core Capabilities

1. Resolve gene → Ensembl ID and disease → EFO/MONDO ID via OT search.
2. Fetch the association (associatedDiseases(Bs:[efo])) with datatypeScores.
3. Report the overall score (strength band) + per-datatype evidence + which datatypes drive it.
4. Flag literature-only links (a target OT may undersell — e.g. somatic/expression-driven antigens).
5. Offline --demo (cached CD276 × lung carcinoma) + reproducibility bundle.

Scope

One skill, one task: the target–disease association evidence breakdown. It does not fetch trials, compute specificity, or score the target overall.

Workflow

1. Resolve (prescriptive): --gene/--ensembl-id and --disease/--efo-id to OT ids (or --demo).
2. Fetch (prescriptive): one GraphQL query for the association row + datatypeScores.
3. Interpret (prescriptive): strength band (strong ≥0.5 / moderate ≥0.1 / weak >0 / none) + drivers; flag literature-only.
4. Emit (prescriptive): association.json + report.md + reproducibility/.

CLI Reference

python skills/opentargets-association-evidence/opentargets_association_evidence.py --gene MET --disease "non-small cell lung carcinoma" --output <dir>
python skills/opentargets-association-evidence/opentargets_association_evidence.py --ensembl-id ENSG00000105976 --efo-id EFO_0003060 --output <dir>
python skills/opentargets-association-evidence/opentargets_association_evidence.py --demo --output <dir>
python clawbio.py run opentargets-association-evidence --gene MET --disease "lung carcinoma"

Demo

python clawbio.py run opentargets-association-evidence --demo

Cached real CD276 × lung carcinoma: overall 0.005 (weak), literature-only — illustrating a somatic/expression-driven antigen that Open Targets' association undersells.

Example Output

association.json (MET × NSCLC, abbreviated):

{
  "symbol": "MET", "disease": "non-small cell lung carcinoma",
  "overall_score": 0.724, "strength": "strong",
  "datatype_scores": {"somatic_mutation": 0.6, "affected_pathway": 0.4, "animal_model": 0.3, "literature": 0.5},
  "interpretation": "association is strong (score 0.724); driven by somatic_mutation, ..."
}

ClawBio is a research and educational tool. It is not a medical device and does not provide clinical diagnoses.

Output Structure

output_directory/
├── association.json          # overall score + datatype_scores + strength + interpretation
├── report.md                 # strength + evidence-by-datatype table
└── reproducibility/
    ├── commands.sh
    ├── environment.yml
    └── checksums.sha256

Dependencies

Required: none beyond Python 3.10+ (OT GraphQL via stdlib urllib). Live mode needs internet; --demo is offline.

Gotchas

• A weak/literature-only association ≠ a bad target. The model will gate on the OT score. Do not — OT under-weights somatic/expression-driven antigens (CD276 × lung is literature-only yet an active ADC target); corroborate with single-cell + clinical skills.
• Datatype scores are not all present. The model will assume all seven datatypes appear. Do not — only datatypes with evidence are returned; absence = no curated evidence, not a zero you can average.
• Disease granularity matters. "lung carcinoma" vs "non-small cell lung carcinoma" vs "small cell lung carcinoma" give different associations. Pass the precise term or --efo-id.
• Scores are harmonic-sum aggregates (0–1), not probabilities. Use the strength band, not the raw number, for conclusions.

Safety

• API-backed, reproducible: official OT GraphQL; no scraping; no fabricated scores.
• Read-only / local-first: writes public data locally.
• Disclaimer: ClawBio is a research and educational tool. It is not a medical device and does not provide clinical diagnoses. Consult a healthcare professional before making any medical decisions.

Agent Boundary

The agent (LLM) picks the target+disease and interprets the evidence profile in context. The skill (Python) performs the API call. The agent must NOT invent datatype scores or treat a low OT association as disqualifying without the other axes.

Chaining Partners

• Complements opentargets-target-factors (target-level) and feeds the Target-ID / Clinical divisions of virtual-biotech-cso. Pairs with clinical-trial-finder (the clinical evidence the OT known_drug datatype reflects). Output is structured JSON, so it chains via the Bio Orchestrator.

Maintenance

• Review cadence: track Open Targets releases (association model + datatype set evolve).
• Staleness signals: associatedDiseases arg/datatypeScores schema changes; new datatypes.
• Deprecation criteria: retire if omics-target-evidence-mapper surfaces the datatype breakdown natively.

Citations

• Open Targets Platform — platform.opentargets.org (target–disease associations, evidence datatypes). API: api.platform.opentargets.org/api/v4/graphql.