MoleculeGNN 🧬

Predicting molecular properties using Graph Neural Networks (GNNs). This project applies deep learning to chemistry by representing molecules as graphs and training neural networks to predict toxicity, solubility, and other properties.

Project Overview

Goal: Build a Graph Neural Network that can predict molecular properties from chemical structure.

Why GNNs?: Traditional machine learning treats molecules as fixed-length feature vectors (fingerprints), losing structural information. GNNs learn directly from the molecular graph structure, capturing complex patterns like functional groups and ring systems.

Architecture

• Input: Molecular structure (SMILES string)
• Representation: Graph (atoms = nodes, bonds = edges)
• Model: Graph Convolutional Network (GCN)
• Output: Physical Property Prediction and Toxicity Prediction for biomolecules

Installation

1. Create a virtual environment and install dependencies:

   python -m venv .venv
   . .venv/bin/activate
   pip install -r requirements.txt

2. Put raw CSV datasets in data/raw/ (example: data/raw/ESOL.csv).

3. Run training:

   python main.py --dataset data/raw/ESOL.csv --model gcn --task regression

Run the app

• Streamlit UI (recommended): From the project root run:

  streamlit run app.py --server.port 8501

  Then open http://localhost:8501 in your browser.

Default model path used by the UI: molecule_gnn_final.pt

Notes

• Modify command-line args in main.py for batch size, epochs, splits, etc.
• Trained model saved to the path given by --save-path.

License

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