DrugPredict

A full-stack bioinformatics application that combines Python-based molecular analysis with a modern React/Next.js web interface for AI-powered drug discovery and compound evaluation.

Quick Start

Prerequisites

• Node.js 18+ and npm
• Python 3.8+ with conda/pip
• Git for version control

Installation

1. Clone the repository

git clone https://github.com/williamhuang3/ml-based-drug-identifier.git
cd ml-based-drug-identifier

2. Set up Python environment

# Install Python dependencies
pip install -r requirements.txt

# Or using conda
conda install -c rdkit rdkit -y
conda install -c conda-forge bash

3. Set up Node.js environment

# Install frontend dependencies
npm install

4. Start the development servers

# Option 1: Start both servers together
npm run dev-full

# Option 2: Start servers separately (two terminals)
npm run flask-dev    # Terminal 1: Flask backend
npm run dev          # Terminal 2: Next.js frontend

5. Open your browser Navigate to http://localhost:3000 (frontend) or http://localhost:5001 (API)

Usage

Web Interface

1. Start the application with npm run dev
2. Enter a target name (e.g., "Coronavirus", "EGFR") or ChemBL ID
3. Click "Search & Analyze" to run the analysis pipeline
4. View results in the organized tabs:
   • Overview: Summary statistics and target information
   • Compounds: Detailed compound data table
   • Statistics: Mann-Whitney U test results
   • Visualizations: Molecular descriptor plots
   • ML Predictions: Random Forest regression results

Command Line (Python)

# Run the Python analysis directly
python main.py

Follow the prompts to:

1. Enter a biological target for analysis
2. Wait for ChemBL data retrieval and processing
3. Run PaDEL descriptor calculation: bash padel.sh
4. View generated plots and statistical results

📊 Analysis Pipeline

1. Target Query: Search ChemBL database for compounds targeting specific proteins
2. Data Preprocessing: Filter and clean compound data, remove duplicates
3. Bioactivity Classification: Label compounds based on IC50 thresholds
4. Molecular Descriptors: Calculate Lipinski descriptors using RDKit
5. Statistical Testing: Perform Mann-Whitney U tests between active/inactive groups
6. Visualization: Generate box plots, scatter plots, and distribution charts
7. Machine Learning: Train Random Forest model using PaDEL descriptors
8. Prediction: Generate IC50 predictions and evaluate model performance

🎯 Key Metrics

• IC50 Classification Thresholds:

  • Active: ≤ 1,000 nM
  • Intermediate: 1,000 - 10,000 nM
  • Inactive: ≥ 10,000 nM

• Lipinski Descriptors:

  • Molecular Weight (MW)
  • Lipophilicity (LogP)
  • Hydrogen Bond Donors
  • Hydrogen Bond Acceptors

• Model Performance Metrics:

  • R² Score (coefficient of determination)
  • RMSE (Root Mean Square Error)
  • MAE (Mean Absolute Error)

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• William Huang - Project Creator
• Data Professor (YouTube) - Inspiration and tutorials
• ChemBL Database - Compound and bioactivity data

---