QUICK_START.md

🧬 Drug Repurposing API - QUICK REFERENCE

🚀 Get Started in 60 Seconds

# Windows
start.bat

# Linux/Mac
chmod +x start.sh && ./start.sh

Wait for "✅ SETUP COMPLETE" message, then visit: http://localhost:8000/docs

---

🎯 Main Endpoint

POST /api/v1/screen

Request:

{
  "disease_name": "Type 2 Diabetes",
  "min_score": 0.5,
  "top_n_targets": 10,
  "known_drugs": ["Metformin"]
}

Response:

{
  "disease": "Type 2 Diabetes",
  "total_targets": 10,
  "total_drugs": 200,
  "total_predictions": 2000,
  "top_results": [
    {
      "drug_name": "Drug_DB00838",
      "target_symbol": "GCK",
      "score": 0.92,
      "status": "✅ Known Treatment"
    }
  ],
  "success": true,
  "message": "✅ Screening completed in 45.23s"
}

---

📊 Other Endpoints

Endpoint	Method	Purpose
/health	GET	Check API status
/api/v1/model-status	GET	Check AI model info
/api/v1/disease-targets	POST	Get disease targets
/api/v1/protein-sequences	POST	Get protein sequences
/api/v1/drug-library	GET	Get drug library

---

🔍 Check Status

curl http://localhost:8000/health
curl http://localhost:8000/api/v1/model-status

---

⚙️ Configuration

Edit app/config.py to adjust:

• MAX_DRUGS_FOR_DEMO: Number of drugs to screen
• MAX_TARGETS: Number of disease targets
• BATCH_SIZE: Optimization for GPU/CPU

---

🆘 Troubleshooting

Problem	Solution
API won't start	Ensure Python 3.10+ installed
DeepPurpose missing	pip install git+https://github.com/kexinhuang12345/DeepPurpose.git
GPU not detected	Install PyTorch CUDA: pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121
Slow predictions	System uses CPU - GPU dramatically faster
No API docs	Visit http://localhost:8000/docs

---

📈 Performance

Config	Speed	Throughput
GPU	~5s	1,200 pairs/sec
CPU	~30s	67 pairs/sec

---

📚 Full Documentation

• PRODUCTION_GUIDE.md - Complete guide
• IMPLEMENTATION_SUMMARY.md - What was built
• http://localhost:8000/docs - Interactive API docs

---

🎓 Pipeline Stages

Disease Input
    ↓
[1] Disease → Targets (OpenTargets API)
    ↓
[2] Targets → Sequences (UniProt API)
    ↓
[3] Load Drug Library (TDC)
    ↓
[4] AI Screening (DeepPurpose MPNN_CNN)
    ↓ GPU CUDA acceleration
    ↓
[5] Process Results
    ↓
Ranked Drug Candidates

---

🔧 Dependencies

Minimum: Python 3.10, pip, 8GB RAM Recommended: GPU with CUDA 12.0+, 16GB RAM

Auto-installed by start scripts:

• FastAPI
• PyTorch
• DeepPurpose (AI model)
• TDC (drug data)

---

Version: 1.0.0 | Status: Production-Ready ✅