tooluniverse-drug-repurposing（工具宇宙药物再利用）

Drug Repurposing with ToolUniverse

Systematically identify and evaluate drug repurposing candidates using multiple computational strategies.

IMPORTANT: Always use English terms in tool calls (drug names, disease names, target names), even if the user writes in another language. Only try original-language terms as a fallback if English returns no results. Respond in the user's language.

Core Strategies

1. Target-Based Repurposing

Start with disease targets → Find drugs that modulate those targets

2. Compound-Based Repurposing

Start with approved drugs → Find new disease indications

3. Disease-Driven Repurposing

Start with disease → Find targets → Match to existing drugs

Quick Start

from tooluniverse import ToolUniverse

tu = ToolUniverse(use_cache=True)
tu.load_tools()

# Example: Find repurposing candidates for a disease
disease_name = "rheumatoid arthritis"

# Step 1: Get disease information
disease_info = tu.tools.OpenTargets_get_disease_id_description_by_name(
    diseaseName=disease_name
)

# Step 2: Get associated targets
disease_id = disease_info['data']['id']
targets = tu.tools.OpenTargets_get_associated_targets_by_disease_efoId(
    efoId=disease_id,
    limit=10
)

# Step 3: Find drugs for each target
for target in targets['data'][:5]:
    drugs = tu.tools.DGIdb_get_drug_gene_interactions(
        gene_name=target['gene_symbol']
    )
    # Evaluate each drug candidate...

Complete Workflow

Phase 1: Disease & Target Analysis

# 1.1 Get disease information
disease_info = tu.tools.OpenTargets_get_disease_id_description_by_name(
    diseaseName="[disease_name]"
)

# 1.2 Find associated targets
targets = tu.tools.OpenTargets_get_associated_targets_by_disease_efoId(
    efoId=disease_info['data']['id'],
    limit=20
)

# 1.3 Get target details for top candidates
target_details = []
for target in targets['data'][:10]:
    details = tu.tools.UniProt_get_entry_by_accession(
        accession=target['uniprot_id']
    )
    target_details.append(details)

Phase 2: Drug Discovery

# 2.1 Find drugs targeting disease-associated targets
drug_candidates = []

for target in targets['data'][:10]:
    # Search DrugBank
    drugbank_results = tu.tools.drugbank_get_drug_name_and_description_by_target_name(
        target_name=target['gene_symbol']
    )

    # Search DGIdb
    dgidb_results = tu.tools.DGIdb_get_drug_gene_interactions(
        gene_name=target['gene_symbol']
    )

    # Search ChEMBL
    chembl_results = tu.tools.ChEMBL_search_drugs(
        query=target['gene_symbol'],
        limit=10
    )

    drug_candidates.extend([drugbank_results, dgidb_results, chembl_results])

# 2.2 Get drug details
for drug_name in unique_drugs:
    # Get DrugBank info
    drug_info = tu.tools.drugbank_get_drug_basic_info_by_drug_name_or_id(
        drug_name_or_drugbank_id=drug_name
    )

    # Get current indications
    indications = tu.tools.drugbank_get_indications_by_drug_name_or_drugbank_id(
        drug_name_or_drugbank_id=drug_name
    )

    # Get pharmacology
    pharmacology = tu.tools.drugbank_get_pharmacology_by_drug_name_or_drugbank_id(
        drug_name_or_drugbank_id=drug_name
    )

Phase 3: Safety & Feasibility Assessment

# 3.1 Check FDA safety data
for drug in top_candidates:
    # Get warnings and precautions
    warnings = tu.tools.FDA_get_warnings_and_cautions_by_drug_name(
        drug_name=drug['name']
    )

    # Get adverse event reports
    adverse_events = tu.tools.FAERS_search_reports_by_drug_and_reaction(
        drug_name=drug['name'],
        limit=100
    )

    # Get drug interactions
    interactions = tu.tools.drugbank_get_drug_interactions_by_drug_name_or_id(
        drug_name_or_id=drug['name']
    )

# 3.2 Assess ADMET properties (for novel formulations)
for drug in top_candidates:
    if 'smiles' in drug:
        admet = tu.tools.ADMETAI_predict_admet(
            smiles=drug['smiles'],
            use_cache=True
        )

Phase 4: Literature Evidence

# 4.1 Search for existing evidence
for drug in top_candidates:
    # PubMed search
    query = f"{drug['name']} AND {disease_name}"
    pubmed_results = tu.tools.PubMed_search_articles(
        query=query,
        max_results=50
    )

    # Europe PMC search
    pmc_results = tu.tools.EuropePMC_search_articles(
        query=query,
        limit=50
    )

    # Clinical trials
    trials = tu.tools.ClinicalTrials_search(
        condition=disease_name,
        intervention=drug['name']
    )

Phase 5: Scoring & Ranking

Create a scoring function to rank candidates:

def score_repurposing_candidate(drug, target_score, safety_data, literature_count):
    """Score drug repurposing candidate (0-100)."""
    score = 0

    # Target association strength (0-40 points)
    score += min(target_score * 40, 40)

    # Safety profile (0-30 points)
    if drug['approval_status'] == 'approved':
        score += 20
    elif drug['approval_status'] == 'clinical':
        score += 10

    if not safety_data.get('black_box_warning'):
        score += 10

    # Literature evidence (0-20 points)
    score += min(literature_count / 5 * 20, 20)

    # Drug-likeness (0-10 points)
    if drug.get('bioavailability') == 'high':
        score += 10

    return score

# Score all candidates
scored_candidates = []
for drug in drug_candidates:
    score = score_repurposing_candidate(
        drug=drug,
        target_score=drug['target_association_score'],
        safety_data=drug['safety_profile'],
        literature_count=drug['supporting_papers']
    )
    drug['repurposing_score'] = score
    scored_candidates.append(drug)

# Sort by score
ranked_candidates = sorted(
    scored_candidates,
    key=lambda x: x['repurposing_score'],
    reverse=True
)

Alternative Strategies

Strategy A: Mechanism-Based Repurposing

# Find drugs with similar mechanism of action
known_drug = "metformin"

# Get mechanism
moa = tu.tools.drugbank_get_drug_desc_pharmacology_by_moa(
    mechanism_of_action="[moa_term]"
)

# Get similar drugs
similar = tu.tools.ChEMBL_search_similar_molecules(
    query=known_drug,
    similarity_threshold=70
)

Strategy B: Network-Based Repurposing

# Use pathway analysis
pathways = tu.tools.drugbank_get_pathways_reactions_by_drug_or_id(
    drug_name_or_drugbank_id="[drug_name]"
)

# Find drugs affecting same pathways
pathway_drugs = tu.tools.drugbank_get_drug_name_and_description_by_pathway_name(
    pathway_name=pathways['data'][0]['pathway_name']
)

Strategy C: Phenotype-Based Repurposing

# Search by indication/phenotype
indication_drugs = tu.tools.drugbank_get_drug_name_and_description_by_indication(
    indication="[related_indication]"
)

# Analyze adverse events as therapeutic effects
# Example: minoxidil (hypertension) → hair growth
adverse_as_therapeutic = tu.tools.FAERS_search_reports_by_drug_and_reaction(
    drug_name="[drug_name]",
    limit=1000
)

Key ToolUniverse Tools

Disease & Target Tools:

• OpenTargets_get_disease_id_description_by_name - Disease lookup
• OpenTargets_get_associated_targets_by_disease_efoId - Disease targets
• UniProt_get_entry_by_accession - Protein details

Drug Discovery Tools:

• drugbank_get_drug_name_and_description_by_target_name - Drugs by target
• drugbank_get_drug_name_and_description_by_indication - Drugs by indication
• DGIdb_get_drug_gene_interactions - Drug-gene interactions
• ChEMBL_search_drugs - Drug search
• ChEMBL_get_drug_mechanisms - Mechanism of action

Drug Information Tools:

• drugbank_get_drug_basic_info_by_drug_name_or_id - Basic drug info
• drugbank_get_indications_by_drug_name_or_drugbank_id - Approved indications
• drugbank_get_pharmacology_by_drug_name_or_drugbank_id - Pharmacology
• drugbank_get_targets_by_drug_name_or_drugbank_id - Drug targets

Safety Assessment Tools:

• FDA_get_warnings_and_cautions_by_drug_name - FDA warnings
• FAERS_search_reports_by_drug_and_reaction - Adverse events
• FAERS_count_death_related_by_drug - Serious outcomes
• drugbank_get_drug_interactions_by_drug_name_or_id - Interactions

Property Prediction Tools:

• ADMETAI_predict_admet - ADMET properties
• ADMETAI_predict_toxicity - Toxicity prediction

Literature Tools:

• PubMed_search_articles - PubMed search
• EuropePMC_search_articles - Europe PMC search
• ClinicalTrials_search - Clinical trials

Output Format

Present results as ranked candidates:

## Drug Repurposing Analysis: [Disease Name]

### Top 10 Repurposing Candidates

#### 1. [Drug Name] (Score: 87/100)

**Current Indications**: [list approved uses]
**Proposed Indication**: [new disease/condition]
**Repurposing Rationale**: Targets [gene/protein] with high association to disease

**Evidence Summary**:
- Target association score: 0.85
- Approval status: FDA approved (safer profile)
- Literature support: 23 papers, 4 clinical trials
- Safety profile: No black box warnings

**Mechanism**: [Brief mechanism description]

**Next Steps**:
- Phase II trial feasibility assessment
- Patient population identification
- Dosing optimization study

**Key Papers**:
1. Smith et al. 2024 - Clinical efficacy in similar condition
2. Jones et al. 2023 - Mechanism validation

---

#### 2. [Drug Name] (Score: 79/100)
[Similar structure...]

### Supporting Analysis

**Target Network**: [visualization or description]
**Pathway Overlap**: [affected pathways]
**Safety Considerations**: [major concerns]
**Development Timeline**: [estimated phases]

Scoring Criteria

Target Association (0-40 points):

• Strong genetic evidence: 40
• Moderate association: 25
• Pathway-level evidence: 15
• Weak/predicted: 5

Safety Profile (0-30 points):

• FDA approved: 20
• Phase III: 15
• Phase II: 10
• Phase I: 5
• No black box warning: +10
• Known serious AE: -10

Literature Evidence (0-20 points):

• Clinical trials: 5 points each (max 15)
• Preclinical studies: 1 point each (max 10)
• Case reports: 0.5 points each (max 5)

Drug Properties (0-10 points):

• High bioavailability: 5
• Good BBB penetration (if CNS): 5
• Low toxicity predictions: 5

Best Practices

1. Start Broad: Query multiple databases (DrugBank, ChEMBL, DGIdb)
2. Validate Targets: Confirm target-disease associations in OpenTargets
3. Check Safety First: Prioritize approved drugs with known safety profiles
4. Literature Mining: Always search for existing clinical/preclinical evidence
5. Use Caching: Enable use_cache=True for expensive predictions
6. Batch Operations: Use tu.run_batch() for parallel queries
7. Consider Mechanism: Evaluate biological plausibility
8. Patent Landscape: Check if indication is already protected
9. Market Analysis: Consider unmet medical need and commercial viability
10. Regulatory Path: FDA approved drugs have faster repurposing path

Common Patterns

Pattern 1: Rapid Screening

# Quick screening of 100+ drugs against disease targets
targets = get_disease_targets(disease_id)[:10]
all_drugs = []

for target in targets:
    drugs = tu.tools.DGIdb_get_drug_gene_interactions(
        gene_name=target['gene_symbol']
    )
    all_drugs.extend(drugs)

# Filter to FDA approved only
approved_drugs = [d for d in all_drugs if d.get('approved')]

Pattern 2: Deep Dive Single Drug

# Comprehensive analysis of one drug candidate
drug_name = "metformin"

# Get everything
info = tu.tools.drugbank_get_drug_basic_info_by_drug_name_or_id(drug_name_or_drugbank_id=drug_name)
targets = tu.tools.drugbank_get_targets_by_drug_name_or_drugbank_id(drug_name_or_drugbank_id=drug_name)
indications = tu.tools.drugbank_get_indications_by_drug_name_or_drugbank_id(drug_name_or_drugbank_id=drug_name)
pharmacology = tu.tools.drugbank_get_pharmacology_by_drug_name_or_drugbank_id(drug_name_or_drugbank_id=drug_name)
interactions = tu.tools.drugbank_get_drug_interactions_by_drug_name_or_id(drug_name_or_id=drug_name)
warnings = tu.tools.FDA_get_warnings_and_cautions_by_drug_name(drug_name=drug_name)
papers = tu.tools.PubMed_search_articles(query=f"{drug_name} AND [new_disease]", max_results=100)

Pattern 3: Comparative Analysis

# Compare multiple candidates side-by-side
candidates = ["drug_a", "drug_b", "drug_c"]

comparison = []
for drug in candidates:
    data = {
        'name': drug,
        'info': tu.tools.drugbank_get_drug_basic_info_by_drug_name_or_id(drug_name_or_drugbank_id=drug),
        'safety': tu.tools.FDA_get_warnings_and_cautions_by_drug_name(drug_name=drug),
        'evidence': tu.tools.PubMed_search_articles(query=drug, max_results=10)
    }
    comparison.append(data)

Troubleshooting

"Disease not found":

• Try disease synonyms or EFO ID lookup
• Use broader disease categories

"No drugs found for target":

• Check target name/symbol (HUGO nomenclature)
• Expand to pathway-level drugs
• Consider similar targets (protein family)

"Insufficient literature evidence":

• Search for drug class rather than specific drug
• Check preclinical/animal studies
• Look for mechanism papers

"Safety data unavailable":

• Drug may not be FDA approved in US
• Check EMA or other regulatory databases
• Review clinical trial safety data

Example Use Cases

Use Case 1: Find repurposing candidates for rare disease

# Rare disease often lack approved drugs
# Strategy: Find drugs targeting same pathways as related common diseases

rare_disease = "Niemann-Pick disease"
related_disease = "Alzheimer's disease"  # Similar pathology

# Get pathways affected in related disease
targets = tu.tools.OpenTargets_get_associated_targets_by_disease_efoId(
    efoId=related_disease_id
)

# Find drugs for those targets
# Evaluate for rare disease applicability

Use Case 2: Repurpose based on adverse effects

# Adverse effect in one context = therapeutic in another
# Example: Thalidomide (teratogenic) → cancer treatment

drug = "drug_name"
adverse_events = tu.tools.FAERS_search_reports_by_drug_and_reaction(
    drug_name=drug,
    limit=1000
)

# Analyze if adverse effects beneficial in other contexts
# Example: weight loss AE → obesity treatment potential

Use Case 3: Combination therapy discovery

# Find drugs that complement existing therapy
primary_drug = "existing_therapy"
disease = "disease_name"

# Get targets not covered by primary drug
disease_targets = tu.tools.OpenTargets_get_associated_targets_by_disease_efoId(
    efoId=disease_id
)

primary_targets = tu.tools.drugbank_get_targets_by_drug_name_or_drugbank_id(
    drug_name_or_drugbank_id=primary_drug
)

# Find drugs for uncovered targets
uncovered_targets = [t for t in disease_targets if t not in primary_targets]

Advanced Techniques

Technique 1: Polypharmacology-Based Repurposing

# Find drugs with multi-target activity matching disease network

# Get disease network
targets = tu.tools.OpenTargets_get_associated_targets_by_disease_efoId(
    efoId=disease_id,
    limit=50
)

# For each drug, count how many disease targets it hits
for drug in candidate_drugs:
    drug_targets = tu.tools.drugbank_get_targets_by_drug_name_or_drugbank_id(
        drug_name_or_drugbank_id=drug
    )

    overlap = len(set(drug_targets) & set(disease_targets))
    if overlap >= 3:  # Multi-target match
        print(f"{drug}: hits {overlap} disease targets")

Technique 2: Structure-Based Repurposing

# Find structurally similar approved drugs

known_active = "known_active_compound"

# Get structure
cid = tu.tools.PubChem_get_CID_by_compound_name(
    compound_name=known_active
)

# Find similar
similar = tu.tools.PubChem_search_compounds_by_similarity(
    cid=cid['data']['cid'],
    threshold=85
)

# Check which are approved drugs
for compound in similar['data']:
    drug_info = tu.tools.PubChem_get_drug_label_info_by_CID(
        cid=compound['cid']
    )

Technique 3: AI-Powered Candidate Selection

# Use ML predictions to filter candidates

candidates_with_smiles = get_candidates_with_structures()

# Predict ADMET for all
admet_results = []
for drug in candidates_with_smiles:
    admet = tu.tools.ADMETAI_predict_admet(
        smiles=drug['smiles'],
        use_cache=True
    )
    admet_results.append({
        'drug': drug['name'],
        'admet': admet,
        'pass': evaluate_admet_criteria(admet)
    })

# Keep only drugs passing ADMET criteria
viable_candidates = [r for r in admet_results if r['pass']]

Resources

For comprehensive disease analysis, see disease-intelligence-gatherer skill.

For compound property analysis, see chemical-compound-retrieval skill.

For detailed ToolUniverse SDK usage, see tooluniverse-sdk skill.