Skip to content

README.md

Latest commit

 

History

History
306 lines (224 loc) · 8.32 KB

README.md

File metadata and controls

306 lines (224 loc) · 8.32 KB

Drug Repurposing Skill

Systematic drug repurposing using ToolUniverse - find new therapeutic uses for existing drugs.

What This Skill Does

This skill enables you to identify, evaluate, and rank drug repurposing candidates using computational approaches. It leverages 750+ scientific tools from ToolUniverse to analyze:

  • Disease-target associations
  • Drug-target interactions
  • Safety profiles
  • Literature evidence
  • Clinical trial data
  • Chemical properties
  • ADMET predictions

When to Use This Skill

Use this skill when you need to:

  • Find repurposing candidates for a specific disease
  • Identify new indications for an approved drug
  • Evaluate repurposing feasibility
  • Assess safety for new indications
  • Mine literature for repurposing evidence
  • Rank multiple candidates systematically

Trigger phrases: "drug repurposing", "drug repositioning", "new indications", "off-label uses", "repurpose drug for", "find drugs for [disease]"

Quick Start

Target-Based Repurposing

Start with disease → Find targets → Match to drugs

from tooluniverse import ToolUniverse

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

# Find drugs for Alzheimer's disease
disease_info = tu.tools.OpenTargets_get_disease_id_description_by_name(
    diseaseName="Alzheimer's disease"
)

targets = tu.tools.OpenTargets_get_associated_targets_by_disease_efoId(
    efoId=disease_info['data']['id'],
    limit=10
)

for target in targets['data'][:3]:
    drugs = tu.tools.DGIdb_get_drug_gene_interactions(
        gene_name=target['gene_symbol']
    )
    # Evaluate candidates...

Compound-Based Repurposing

Start with drug → Find targets → Match to diseases

# Find new uses for metformin
drug_targets = tu.tools.drugbank_get_targets_by_drug_name_or_drugbank_id(
    drug_name_or_drugbank_id="metformin"
)

for target in drug_targets['data'][:5]:
    diseases = tu.tools.OpenTargets_get_diseases_by_target_ensemblId(
        ensemblId=target['ensembl_id']
    )
    # Check if new indication...

Files in This Skill

  • SKILL.md - Main skill instructions and workflows
  • EXAMPLES.md - 7 complete worked examples
  • REFERENCE.md - Detailed tool documentation
  • README.md - This file

Repurposing Strategies

This skill supports multiple approaches:

1. Target-Based

Disease → Targets → Drugs → Validation

Best for: Diseases with known molecular targets

2. Compound-Based

Drug → Targets → Diseases → Validation

Best for: Finding new uses for approved drugs

3. Disease-Driven

Disease → Targets → Pathways → Drugs

Best for: Complex/polygenic diseases

4. Mechanism-Based

Known MOA → Similar MOA drugs → Evaluation

Best for: Mechanism-validated hypotheses

5. Network-Based

Pathway analysis → Pathway overlap → Candidate drugs

Best for: Systems biology approaches

6. Phenotype-Based

Adverse events → Therapeutic potential → Validation

Best for: Mining unexpected effects

7. Structure-Based

Active compound → Similar structures → Approved analogs

Best for: SAR-driven repurposing

Key Features

Comprehensive Data Integration

  • 10+ databases: DrugBank, ChEMBL, OpenTargets, PubChem, FDA, FAERS, PubMed, etc.
  • Multiple evidence types: Genetic, chemical, clinical, literature
  • Cross-validation: Verify across multiple sources

Systematic Scoring

  • Target association strength
  • Safety profile assessment
  • Literature evidence quantification
  • Drug-likeness evaluation
  • Overall repurposing potential (0-100)

Safety Assessment

  • FDA warnings and precautions
  • FAERS adverse event analysis
  • Drug-drug interactions
  • Risk stratification

Evidence Mining

  • PubMed literature search
  • Clinical trial identification
  • Mechanism validation
  • Patent landscape

Output Format

All analyses produce ranked candidate lists:

## Top Repurposing Candidates

1. Drug Name (Score: 87/100)
   - Target: GENE (association: 0.85)
   - Status: FDA approved
   - Evidence: 23 papers, 4 trials
   - Safety: No black box warnings
   - Recommendation: Immediate Phase II planning

2. Drug Name (Score: 79/100)
   ...

Examples Overview

The skill includes 7 comprehensive examples:

  1. Target-Based for Alzheimer's - Complete workflow from disease to candidates
  2. Compound-Based for Metformin - Finding new indications for approved drug
  3. Disease-Driven for COVID-19 - Emergency repurposing rapid screening
  4. Network-Based Pathway Analysis - Using pathway overlap
  5. Structure-Based Repurposing - Finding approved analogs
  6. Adverse Event Mining - Converting AEs to therapeutic uses
  7. Multi-Database Integration - Comprehensive scoring system

Each example is fully executable with ToolUniverse.

Prerequisites

Installation

pip install tooluniverse

Optional Features

pip install tooluniverse[ml]  # For ADMET predictions

Environment

export OPENAI_API_KEY="sk-..."  # For LLM-based tool search
export NCBI_API_KEY="..."       # For higher PubMed rate limits

Performance Tips

  1. Enable caching: tu = ToolUniverse(use_cache=True)
  2. Limit initial searches: Start with top 10-20 results
  3. Use batch operations: Parallel queries when possible
  4. Filter early: Apply approval status filter first
  5. Validate top candidates: Deep dive on highest scorers only

Common Workflows

Quick Screening (15 minutes)

  1. Get disease targets (top 10)
  2. Find drugs for each target
  3. Filter to FDA approved
  4. Rank by evidence count
  5. Report top 5

Comprehensive Analysis (2-4 hours)

  1. Disease and target analysis
  2. Multi-database drug discovery
  3. Safety assessment (FDA + FAERS)
  4. Literature mining (PubMed + trials)
  5. ADMET predictions
  6. Scoring and ranking
  7. Detailed report with recommendations

Emergency Repurposing (1 hour)

  1. Rapid target identification
  2. Approved drug screening
  3. Safety filtering
  4. Evidence scoring
  5. Priority recommendations

Integration with Other Skills

This skill works well with:

  • disease-intelligence-gatherer - Comprehensive disease analysis first
  • chemical-compound-retrieval - Detailed compound property analysis
  • literature-deep-research - Systematic literature reviews
  • protein-structure-retrieval - Target structure for rational design

Validation Checklist

Before recommending repurposing candidate:

  • Target-disease association validated (score > 0.7)
  • Drug approval status confirmed
  • Safety profile acceptable for indication
  • Literature evidence reviewed
  • Mechanism biologically plausible
  • No contraindications for patient population
  • Patent landscape checked
  • Market need assessed
  • Clinical trial feasibility considered
  • Regulatory pathway identified

Success Metrics

High Potential Candidates (Score ≥70):

  • Strong target association (>0.8)
  • FDA approved status
  • Substantial literature (>20 papers)
  • Clinical trials ongoing/completed
  • Favorable safety profile

Moderate Potential (Score 50-69):

  • Moderate target association (0.5-0.8)
  • Phase II/III status acceptable
  • Some literature evidence (5-20 papers)
  • Manageable safety concerns

Low Potential (Score <50):

  • Weak association (<0.5)
  • Limited evidence (<5 papers)
  • Safety concerns
  • Requires more validation

Limitations

  1. Computational predictions: Not clinical proof
  2. Database coverage: Not all drugs/targets included
  3. Literature bias: Publication bias affects evidence
  4. Safety data: Historical, may not reflect all risks
  5. Mechanism assumptions: Computational models have limits

Always: Validate top candidates through clinical trials

Citation

When using this skill, cite:

Support

Version

  • Version: 1.0.0
  • Last Updated: February 2026
  • Compatible with: ToolUniverse 0.5+

License

This skill follows ToolUniverse licensing. Check individual database terms of use for commercial applications.