Scientific Drug Target Profiling

Phase 1: Identity Resolution

ID マッピングチェーン

def resolve_target_ids(query):
    """
    ターゲット名/Gene Symbol から主要 ID を解決する。
    """
    ids = {
        "query": query,
        "uniprot_accession": None,     # e.g., P04637
        "ensembl_id": None,            # e.g., ENSG00000141510
        "entrez_id": None,             # e.g., 7157
        "chembl_target_id": None,      # e.g., CHEMBL3927
        "hgnc_symbol": None,           # e.g., TP53
        "open_targets_id": None,       # = Ensembl ID
    }

    # Step 1: UniProt 検索（Gene Name → Accession）
    # Step 2: Ensembl ID 取得（UniProt xref）
    # Step 3: ChEMBL Target ID（UniProt → ChEMBL mapping）
    # Step 4: Cross-validation（各 DB 間で一致確認）

    return ids

名前衝突の検出: 同名の遺伝子が複数生物種に存在する場合、 UniProt taxonomy filter で human (9606) を優先する。

Phase 2: Druggability Assessment

ドラッガビリティ 3 軸評価

┌─────────────────────────────────────────┐
│         Druggability Matrix             │
├─────────────┬───────────┬───────────────┤
│  Modality   │  Metric   │   Threshold   │
├─────────────┼───────────┼───────────────┤
│ Small Mol   │ Pocket?   │ ≥1 druggable  │
│ Antibody    │ Surface?  │ extracellular │
│ PROTAC      │ E3 dist   │ ≤30 Å         │
│ ASO/siRNA   │ mRNA expr │ detectable    │
│ Gene Therapy│ LOF/GOF   │ disease link  │
└─────────────┴───────────┴───────────────┘

Target Development Level (TDL) 分類

def classify_tdl(target_data):
    """
    Pharos TDL 分類に準拠したターゲット分類。
    Tclin: 承認薬あり
    Tchem: 高活性化合物あり（ChEMBL）
    Tbio: 生物学的機能が判明
    Tdark: 情報不足
    """
    if target_data.get("approved_drugs"):
        return "Tclin"
    elif target_data.get("chembl_activities_count", 0) > 0:
        potent = [a for a in target_data["activities"]
                  if a.get("pchembl_value", 0) >= 6.0]
        if potent:
            return "Tchem"
    if target_data.get("go_annotations") or target_data.get("publications", 0) > 5:
        return "Tbio"
    return "Tdark"

Phase 3: Safety Profiling

安全性評価チェックリスト

## Safety Assessment

### Genetic Constraint
- [ ] pLI score: ___  (>0.9 = highly constrained, LOF intolerant)
- [ ] LOEUF: ___  (<0.35 = constrained)
- [ ] Missense Z-score: ___  (>3.09 = missense-constrained)

### Essential Gene Analysis
- [ ] DepMap dependency score: ___  (<-0.5 = broadly essential)
- [ ] Mouse knockout phenotype: ___
- [ ] Lethal phenotype: ___  (YES/NO)

### Expression Breadth
- [ ] Tissue specificity index (tau): ___
- [ ] Ubiquitously expressed: ___  (risk for on-target toxicity)
- [ ] Brain/Heart/Liver expression: ___  (safety-critical organs)

### Off-target Risk
- [ ] Paralog count: ___
- [ ] Closest paralog similarity: ___  %
- [ ] Shared binding site features: ___

Phase 4: Disease Association

エビデンスグレーディング

EVIDENCE_TIERS = {
    "T1": "Genetic + Clinical (GWAS + ClinVar pathogenic)",
    "T2": "Strong biological (functional studies + animal models)",
    "T3": "Associative (expression correlation + network guilt-by-association)",
    "T4": "Computational prediction only",
}

def grade_disease_association(target_id, disease_id, evidence_sources):
    """
    ターゲット-疾患アソシエーションのエビデンス評価。
    Open Targets overall_association_score + 追加エビデンスで T1-T4 判定。
    """
    score = evidence_sources.get("open_targets_score", 0)
    has_gwas = evidence_sources.get("gwas_significance", False)
    has_clinvar = evidence_sources.get("clinvar_pathogenic", False)
    has_functional = evidence_sources.get("functional_study", False)

    if has_gwas and has_clinvar:
        return "T1", score
    elif has_functional or score > 0.7:
        return "T2", score
    elif score > 0.3:
        return "T3", score
    else:
        return "T4", score

Phase 5: Competitive Landscape

パイプラインマッピング

## Competitive Intelligence

### Known Drugs (Approved)
| Drug | Mechanism | Indication | Approval Year |
|------|-----------|------------|---------------|

### Clinical Pipeline
| Compound | Phase | Sponsor | Indication | NCT ID |
|----------|-------|---------|------------|--------|

### Chemical Probes
| Probe | Potency | Selectivity | Source |
|-------|---------|-------------|--------|

### Patent Landscape
| Patent Family | Assignee | Filing Date | Key Claims |
|---------------|----------|-------------|------------|

Report Template

ターゲットインテリジェンスレポート

# Target Intelligence Report: [TARGET NAME]

**Generated**: [Date] | **Analyst**: SATORI Drug Target Profiling

## 1. Executive Summary
[2-3 sentences: target name, key disease links (with evidence tier), druggability verdict]

## 2. Target Identifiers
| Database | ID | Verified |
|----------|----|----------|
| UniProt  |    | ✓/✗     |
| Ensembl  |    | ✓/✗     |
| ChEMBL   |    | ✓/✗     |

## 3. Protein Biology
### 3.1 Function & Localization
### 3.2 Domain Architecture
### 3.3 Pathway Context

## 4. Structural Biology
### 4.1 Experimental Structures (PDB)
### 4.2 AlphaFold Prediction
### 4.3 Binding Sites & Pockets

## 5. Expression Profile
### 5.1 Tissue Expression (GTEx/HPA)
### 5.2 Disease-specific Expression

## 6. Disease Associations
[Table with evidence tiers T1-T4]

## 7. Druggability Assessment
### 7.1 TDL Classification
### 7.2 Modality Assessment
### 7.3 Tractability Score

## 8. Known Ligands & Drugs
### 8.1 Approved Drugs
### 8.2 Clinical Candidates
### 8.3 Chemical Probes & Tool Compounds

## 9. Safety Profile
### 9.1 Genetic Constraint
### 9.2 Essential Gene Status
### 9.3 Off-target Risk

## 10. Competitive Landscape

## 11. Recommendations
### 11.1 Go/No-Go Assessment
### 11.2 Suggested Modality
### 11.3 Key Experiments Needed

## 12. Data Sources & Methodology

Completeness Checklist

必須項目

• ID Resolution: UniProt, Ensembl, ChEMBL, HGNC の 4ID 確認
• Druggability: TDL 分類 + 少なくとも 2 モダリティの評価
• Safety: pLI + LOEUF + DepMap の 3 指標
• Disease: 上位 5 疾患の T1-T4 グレード
• Literature: 主要レビュー論文 ≥3 件引用
• Competitive: 承認薬と Ph3 候補を網羅

Best Practices

1. ID は必ず Cross-validate: UniProt と Ensembl の双方向マッピングで確認
2. 名前衝突を検出: Gene Symbol の同名異義語を taxonomy filter で排除
3. Evidence Tier を明記: すべての疾患アソシエーションに T1-T4 を付与
4. 安全性を最初に確認: Essential gene の場合は早期に Go/No-Go 判断
5. 構造情報を優先: PDB 実験構造 > AlphaFold 予測 > Homology model

References

Output Files

利用可能ツール

ToolUniverse SMCP 経由で利用可能な外部ツール。

参照スキル