Molecular Property Predictor

A scalable Flask application for predicting molecular properties from SMILES strings using Random Forest regression with robust SMILES validation and molecular structure visualization.

Features

• Robust SMILES Validation: Multi-level validation using RDKit built-in functions

  • Parse validation
  • Atom count check
  • Valence validation
  • Radical electron detection

• Molecular Structure Visualization: 2D structure rendering with heavy atoms only (hydrogens removed)

• Property Prediction: Random Forest models for:

  • Molecular Weight
  • LogP (lipophilicity)
  • TPSA (Topological Polar Surface Area)

Project Structure

molecular_predictor/
├── app.py                  # Main Flask application
├── config.py              # Configuration settings
├── models/
│   └── predictor.py       # ML model implementation
├── utils/
│   └── validators.py      # SMILES validation & image generation
├── static/
│   ├── css/
│   │   └── style.css     # Stylesheets
│   └── js/
│       └── main.js       # Frontend JavaScript
├── templates/
│   └── index.html        # HTML template
├── requirements.txt       # Python dependencies
└── README.md             # This file

Installation

1. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

Usage

Run the application:

python app.py

Visit http://localhost:5000 in your browser.

SMILES Validation

The app performs comprehensive validation:

1. Format Check: Ensures SMILES is a non-empty string
2. Parsing: Validates molecular structure can be parsed
3. Sanitization: Checks for valid valence states
4. Quality Checks: Detects radicals and unusual structures

API Endpoints

• GET / - Main web interface
• POST /predict - Predict molecular properties
  • Request: {"smiles": "CCO"}
  • Response: {"smiles": "CCO", "canonical_smiles": "CCO", "predicted_properties": {...}, "molecule_image": "base64..."}
• GET /health - Health check endpoint

Example SMILES to Try

Valid:

• CCO - ethanol
• c1ccccc1 - benzene
• CC(=O)O - acetic acid
• CC(C)(C)C - neopentane

Invalid (will show specific error):

• XYZ - Invalid characters
• C[ - Incomplete structure
• C=C=C=C - Valence issues

Configuration

Edit config.py to modify:

• Model parameters (n_estimators, fingerprint_bits)
• Server settings (host, port, debug mode)
• Environment-specific configurations

Future Enhancements

• Add model training endpoint
• Implement batch predictions
• Add user authentication
• Deploy with Docker
• Add database for storing predictions
• Support for 3D structure visualization