GraphPINE: Graph Propagating Importance Network for Explanation

This is the official implementation for GraphPINE: Enhancing Interpretable Graph Neural Networks with Prior Knowledge through Importance Propagation.

GraphPINE is a novel Graph Neural Network (GNN) designed to provide interpretable predictions of drug response through the propagation of gene importance scores. This project integrates multi-omics data with graph structural information to enhance the prediction accuracy and interpretability of drug-gene interactions, offering significant advancements in computational biology, personalized medicine, and drug discovery.

Features

• Interpretable Predictions: GraphPINE incorporates an innovative ImportancePropagationLayer, which updates and propagates gene importance scores across the network during training, allowing for detailed insights into drug mechanisms.
• Multi-omics Integration: The model uses gene expression, copy number variation, methylation, and mutation data to build a comprehensive gene-gene interaction network.
• Advanced GNN Architecture: GraphPINE leverages state-of-the-art GNN layers such as GAT, GINE, and Graph Transformer to process graph data with edge attributes.
• Scalable and Flexible: The model is designed to handle large-scale biological datasets and can be adapted for various drug response prediction tasks.

Quick Start Guide

This quick start guide provides instructions on how to run GraphPINE predictions using both CPU and GPU, with the process completing in just a few seconds.

Steps:

1. Download the Repository:
   Begin by downloading the repository to your local machine.

2. Unzip the Repository:
   Use the command unzip [REPOSITORY_DIRECTORY].zip to extract the files.

3. Change Directory:
   Navigate into the repository directory with cd [REPOSITORY_DIRECTORY].

4. Build the Docker Image:
   Build your Docker image using docker build -t [YOUR_DOCKER_USERNAME]/[YOUR_IMAGE_NAME] ..

5. Run the Docker Container:
   Start the Docker container using docker run -it -p 9999:9999 [YOUR_DOCKER_USERNAME]/[YOUR_IMAGE_NAME].

   After starting the Docker container, access the Jupyter Notebook by navigating to http://localhost:9999/notebooks/Tutorial-pretrained_model.ipynb in your web browser and run all cells. This notebook will guide you through the basic usage of GraphPINE and demonstrate example predictions.

Requirements

accelerate==0.33.0
numpy==1.26.4
pandas==2.2.2
matplotlib==3.9.2
scikit-learn==1.2.2
torch==2.4.0
torch_geometric==2.5.3
torch_scatter==2.1.2
torchaudio==2.4.0
torchvision==0.19.0
tqdm==4.66.5
requests==2.32.3
plotly==5.23.0
networkx==3.3
seaborn==0.13.2