tooluniverse-target-research

What this skill does

# Comprehensive Target Intelligence Gatherer

Gather complete target intelligence by exploring 9 parallel research paths. Supports targets identified by gene symbol, UniProt accession, Ensembl ID, or gene name.

**KEY PRINCIPLES**:
1. **Report-first approach** - Create report file FIRST, then populate progressively
2. **Tool parameter verification** - Verify params via `get_tool_info` before calling unfamiliar tools
3. **Evidence grading** - Grade all claims by evidence strength (T1-T4)
4. **Citation requirements** - Every fact must have inline source attribution
5. **Mandatory completeness** - All sections must exist with data minimums or explicit "No data" notes
6. **Disambiguation first** - Resolve all identifiers before research
7. **Negative results documented** - "No drugs found" is data; empty sections are failures
8. **Collision-aware literature search** - Detect and filter naming collisions
9. **English-first queries** - Always use English terms in tool calls, even if the user writes in another language. Translate gene names, disease names, and search terms to English. Only try original-language terms as a fallback if English returns no results. Respond in the user's language

---

## LOOK UP, DON'T GUESS

When asked about a specific protein or gene target, look it up in UniProt/Ensembl/OpenTargets BEFORE reasoning about it. Verify the gene name, function, and disease associations from databases. When you're not sure about a fact, your first instinct should be to SEARCH for it using tools, not to reason harder from memory.

---

## When to Use This Skill

Apply when users:
- Ask about a drug target, protein, or gene
- Need target validation or assessment
- Request druggability analysis
- Want comprehensive target profiling
- Ask "what do we know about [target]?"
- Need target-disease associations
- Request safety profile for a target

**When NOT to use**: Simple protein lookup, drug-only queries, disease-centric queries, sequence retrieval, structure download — use specialized skills instead.

---

## Target Evaluation Reasoning Framework

Evaluating a drug target requires reasoning across four interconnected questions. Answer all four before forming a recommendation.

**1. Is there genetic evidence linking this target to the disease?**
Genetic evidence is the strongest predictor of drug success — targets with human genetic support have approximately twice the clinical success rate as those without (Nelson et al. 2015). Ask: Are there GWAS associations connecting this gene to the disease? Do rare loss-of-function or gain-of-function variants cause or protect against the disease? Does the mouse knockout phenotype match the human disease (from OpenTargets mouse models)? OpenTargets assigns genetic evidence scores; a score > 0.7 indicates strong support. ClinVar rare variant evidence and DisGeNET curated gene-disease association scores add complementary layers. A target with no genetic link to the disease of interest carries a fundamental validation risk that cannot be resolved by downstream data.

**2. Is the target druggable?**
Druggability has two components: structural accessibility and prior chemical matter. Structural accessibility means the target has a binding pocket where a small molecule or biologic can engage — surface-exposed receptors, enzymes with well-defined active sites, and protein-protein interaction interfaces with hot spots are tractable. Intrinsically disordered proteins and transcription factors with flat, featureless binding surfaces are typically harder. Pharos TDL classification provides a tiered assessment: Tclin (approved drug), Tchem (known active compounds), Tbio (biological function known but no drugs), Tdark (poorly characterized). If ChEMBL or BindingDB have compounds with IC50 < 1μM, the target is chemically tractable. Chemical probes (from OpenTargets chemical probes endpoint) confirm a target can be modulated, which is distinct from drug-like compounds. For GPCRs, check GPCRdb for curated agonists and antagonists.

**3. Is the target safe to modulate?**
Safety concerns arise from two sources. First, on-target effects: if the target is essential in normal tissues (mouse KO is lethal, or gnomAD pLI is high / LOEUF is low), full inhibition will produce toxicity — the question becomes whether a partial agonist or tissue-targeted delivery can provide a therapeutic window. Second, off-target effects: does the gene have family members that could be inadvertently hit? The OpenTargets safety profile aggregates known toxicity annotations, and DepMap essentiality scores tell you which cancer cell lines require this gene for survival (useful but not directly translatable to normal tissues). Expression specificity matters: a target expressed only in the disease-relevant tissue is far safer than one expressed ubiquitously in critical organs (heart, kidney, brain).

**4. What is the competitive landscape?**
A target with approved drugs may already be validated but competitive; a target with clinical-stage programs from competitors establishes feasibility while creating IP barriers. An entirely novel target with no drug history requires more extensive internal validation. Assess: number of ChEMBL bioactivity records (chemical matter depth), approved drugs from OpenTargets drug associations, and literature activity trends (recent paper count and key research groups). A dark target (Tdark) with strong genetic evidence but no chemical matter is a high-risk, high-reward opportunity.

**Synthesizing the four dimensions**: The ideal target has strong genetic evidence (GWAS + rare variant), a tractable binding site (Tclin or Tchem), acceptable safety profile (tissue-specific expression, non-lethal KO), and manageable competition. Gaps in any dimension represent validation tasks, not disqualifiers — but they must be acknowledged. A target with perfect druggability but no genetic link to disease is a tractability exercise, not a validated therapeutic hypothesis.

---

## Phase 0: Tool Parameter Verification (CRITICAL)

**BEFORE calling ANY tool for the first time**, verify its parameters:

```python
tool_info = tu.tools.get_tool_info(tool_name="Reactome_map_uniprot_to_pathways")
# Reveals: takes `id` not `uniprot_id`
```

Known parameter corrections:
- `Reactome_map_uniprot_to_pathways`: param is `id` (not `uniprot_id`)
- `ensembl_get_xrefs`: param is `id` (not `gene_id`)
- `GTEx_get_median_gene_expression`: requires `gencode_id` + `operation="median"`; try versioned Ensembl ID if empty
- `OpenTargets_*`: param is `ensemblId` (camelCase, not `ensemblID`)
- `STRING_get_protein_interactions`: takes `protein_ids` (list) + `species`
- `intact_get_interactions`: takes `identifier` (UniProt accession, not gene symbol)

---

## Critical Workflow Requirements

**Report-First (MANDATORY)**: Create `[TARGET]_target_report.md` with all section headers and `[Researching...]` placeholders before starting research. Update progressively. Do not show raw tool outputs to the user.

**Evidence Grading (MANDATORY)**: Grade every claim T1-T4. T1 = clinical/genetic data; T2 = curated databases or multiple studies; T3 = computational or single study; T4 = annotation or catalog entry.

---

## Core Strategy: 9 Research Paths

```
Target Query (e.g., "EGFR" or "P00533")
|
+- IDENTIFIER RESOLUTION (always first)
|   +- Check if GPCR -> GPCRdb_get_protein
|
+- PATH 0: Open Targets Foundation (ALWAYS FIRST - fills gaps in all other paths)
|
+- PATH 1: Core Identity (names, IDs, sequence, organism)
|   +- InterProScan_scan_sequence for novel domain prediction
+- PATH 2: Structure & Domains (3D structure, domains, binding sites)
|   +- If GPCR: GPCRdb_get_structures (active/inactive states)
+- PATH 3: Function & Pathways (GO terms, pathways, biological role)
+- PATH 4: Protein Interactions (PPI network, complexes)
+- PATH 5: Expression Profile (tissue expression, single-cell)
+- PATH 6: Variants & Disease (mutations, clinical significance)
|   +- DisGeNET_sea