| name | score-target |
| description | Deep-dive validation of a single gene target across one or more diseases. Use when the user wants detailed scoring, a deep-dive, SWOT analysis, or comprehensive profile of a single drug target. Triggers on "score target", "deep dive", "profile target", "analyze target". |
Deep-Dive Target Scoring
Perform an in-depth validation of a single target across one or more diseases.
Usage
/score-target PCSK9 — cardiovascular disease, familial hypercholesterolemia
Instructions
Given $ARGUMENTS:
-
Parse. Extract one target and one or more diseases from the input.
-
Resolve. Use the open-targets-platform MCP tool search_entities to map to
canonical Ensembl Gene IDs and EFO Disease IDs.
-
Deep data collection. Go beyond what open_targets_client.py fetches — use the
MCP tools get_open_targets_graphql_schema/get_type_dependencies to find the right
fields, then query_open_targets_graphql (or batch_query_open_targets_graphql for
multiple diseases at once) to also fetch:
- Genetic constraint (pLI, LOEUF from gnomAD via Open Targets)
- Mouse phenotypes (IMPC data from Open Targets)
- Tissue expression (baseline expression from Open Targets)
- Pharmacogenomics data if available
- Safety liabilities and adverse events
-
Supplementary search. Use web search to find:
- Latest clinical trial updates on ClinicalTrials.gov
- Recent publications (last 2 years) about this target
- Patent landscape (any composition-of-matter patents?)
- Competitive landscape (who else is developing drugs against this target?)
-
Score. Apply the standard rubrics, but also provide sub-scores:
- Clinical: Genetic evidence sub-score + Trial evidence sub-score
- Druggability: Structural sub-score + Chemical matter sub-score + Modality sub-score
- Pathway: Pathway centrality sub-score + Expression sub-score + Animal model sub-score
-
Output. Present a detailed profile card for the target including:
- Summary table with all scores
- Strengths / Weaknesses / Opportunities / Threats (SWOT) analysis
- Comparison across diseases if multiple provided
- Recommended next experiments
-
Generate HTML report. Run the Python script to produce the interactive HTML matrix:
python open_targets_client.py validate "TARGET" "DISEASE1,DISEASE2"
This creates results/validation_{disease}.html — the interactive scoring matrix.
Tell the user to open it in their browser.
-
Save narrative to results/{TARGET}_deep_dive.md alongside the HTML.