PPI-OMEGA

This is the project repository for Data Science Capstone at UCSD in 2025 by Team A06-1.

Protein-Protein Interaction with Omics-Enhanced Graph Autoencoder, a.k.a. PPI-OMEGA, is a Variational Graph Autoencoder (VGAE)-based framework designed to improve Protein-Protein Interaction (PPI) predictions by integrating multi-omics data. Unlike traditional models that rely solely on static network topology, PPI-OMEGA incorporates RNA expression profiles and protein expression data to learn biologically meaningful representations.

---

Installation and Execution Guide

Option 1: Running with Docker (Recommended)

If you prefer a pre-configured environment, you can use Docker.

1. Install Docker

Ensure you have Docker installed. You can download it from here.

2. Pull the Docker Image

You can pull the pre-built Docker image directly (if it's available on Docker Hub):

docker pull eliteapex/ppi-omega

Alternatively, you can build the image manually:

git clone https://github.com/EliteApex/PPI-OMEGA.git
cd PPI-OMEGA
docker build -t ppi-omega .

3. Run the Docker Container

Run the container interactively:

docker run -it --rm -v $(pwd):/app ppi-omega bash

This will mount your current directory (PPI-OMEGA) inside the container, so you can access scripts and data.

4. Run the Model inside Docker

Inside the container:

python src/run_models.py --version <version_num>

where <version_num> = 1, 2, or 3 depending on the input features you'd like to use.

5. Using VS Code with Docker

To use VS Code with the Docker container:

1. Install the Remote - Containers extension.
2. Open VS Code and connect to the container:
   • Open Command Palette (Ctrl+Shift+P).
   • Select Remote-Containers: Attach to Running Container.
   • Choose ppi-omega from the list.
3. You can now use VS Code as if working in a local environment.

---

Option 2: Running Locally with Conda

If you don't want to use Docker, you can manually set up the environment.

1. Clone the Repository

git clone https://github.com/EliteApex/PPI-OMEGA.git
cd PPI-OMEGA

2. Set Up Conda Environment

Ensure that Conda is installed. Then, create and activate the environment:

conda env create -f environment.yml
conda activate PPIOMEGA_env  

3. Run the Model

Once the environment is set up, you can run the model:

python src/run_models.py --version <version_num>

or within a Jupyter Notebook:

%run src/run_models.py --version <version_num>

where <version_num> = 1, 2, or 3 depending on the input features you'd like.

---

Project Structure

The repository is organized as follows:

.
├── Data/                      # Dataset directory storing original and intermediate data files
│   ├── raw/                   # Raw data files
│   │   ├── normal_ihc_data.tsv
│   │   ├── protein_gene_conversion.csv
│   │   ├── rna_tissue_gtex.tsv
│   ├── adj_matrix_scaled.npz
│   ├── adj_matrix.npz
│   ├── filtered_PPI.csv
│   ├── PPI_protein_expression_full.csv
│   ├── PPI_Protein_only.csv
│   ├── PPI_RNA_only.csv
│   ├── PPI_RNA_Protein_combined.csv
│   ├── PPI_RNA_seq_full.csv
│   ├── protein_gene_conversion.csv
│   ├── protein_node_id_conversion.csv
│   └── protein_vis_samples.csv
├── notebooks/                 # Jupyter notebooks for analysis and visualization
│   ├── EDA.ipynb
├── plots/                     # Directory for storing plots and visualizations
├── scripts/                   # Additional scripts for data processing
│   ├── preprocessing.py
│   ├── vis_ppi_network_sample_data.py
├── src/                       # Source code for the project
│   ├── _pycache__/            # Cached Python files
│   ├── baseline_model.py
│   ├── best_hyperparameters.csv
│   ├── best_model.pth
│   ├── latent_parameters_v0.csv
│   ├── latent_parameters_v1.csv
│   ├── latent_parameters_v2.csv
│   ├── latent_parameters_v3.csv
│   ├── latent_variables_sampled.csv
│   ├── metrics_version_0.npz
│   ├── metrics_version_1.npz
│   ├── metrics_version_2.npz
│   ├── metrics_version_3.npz
│   ├── model.py
│   ├── pilot_with_features.ipynb
│   ├── run_models.py
│   ├── selected_nodes.txt
│   ├── vgae_cv.py
│   └── visualization.ipynb
├── .dockerignore              # Files and directories to ignore in Docker builds
├── .gitignore                 # Git ignore rules
├── Dockerfile                 # Docker container definition
├── environment.yml            # Conda environment file
└── README.md                  # Documentation

---

Contributors

• Team A06-1 - UCSD Data Science Capstone 2025
  • Siddharth Vyasabattu
  • Xiaoyu Gui

---