MolGen 🧬

AI-Powered Drug Molecule Generation System

MolGen is a full-stack application that generates drug-like molecules from natural language prompts, evaluates them, and returns the best candidates in real time.

---

🚀 One-Line Pitch

Describe desired drug properties → get ranked, chemically valid molecules instantly.

---

🧠 Why This Matters

Drug discovery is costly (~$2.6B per drug) and slow (~12 years) with high failure rates.

MolGen reduces this by:

• Generating candidates computationally
• Evaluating properties before lab testing
• Ranking the best molecules automatically

---

⚙️ How It Works

🔁 Pipeline

Prompt → ML API → Generate Molecules → Score → Rank → Top Results

---

🧩 System Architecture

Layer	Component	Description
Generation	ML API	Generates molecules from prompts
Backend	FastAPI	Handles API + scoring
Frontend	Next.js	Displays results
Database	MongoDB Atlas	Stores generated molecules

---

🔬 Key Features

• 🔁 Generate-then-filter pipeline
• 📊 Property scoring (QED, LogP, TPSA)
• 🧪 RDKit validation
• 💾 MongoDB storage
• 🌐 Full-stack deployment

---

📊 Example Output

Molecule (SMILES)	QED	Lipinski	TPSA
CCOC1=CC=CC=C1	0.82	✅	42.3
CCN(CC)CCO	0.76	✅	35.1

---

🛠 Tech Stack

• Backend: Python, FastAPI
• Frontend: Next.js, Tailwind
• DB: MongoDB Atlas
• Deployment: Render, Vercel

---

📸 Demo

prompt input

result

---

⚙️ Setup

Backend

python -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r backend/requirements.txt
cd backend
..\.venv\Scripts\python.exe -m uvicorn main:app --reload

Frontend

cd frontend
npm install
npm run dev

📡 API

POST /generate → returns:

• SMILES
• QED
• Lipinski
• TPSA

---

⚠️ Limitations

• Computational predictions (not lab-tested)
• Depends on API/model quality