This project implements a pipeline to analyze influenza virus sequences and predict potential mutations using a Codon-based Graph Neural Network (GNN). The pipeline:

• Reads metadata from a CSV file and sequences from a FASTA file.
• Normalizes and filters metadata based on available sequences.
• Parses collection dates and selects a baseline sequence (earliest date).
• Performs pairwise sequence alignment (demo).
• Converts nucleotide sequences into codon-based graphs (nodes are one-hot encoded codons with sequential edges).
• Trains a Graph Convolutional Network (GCN) to predict mutation probabilities (dummy labels in this demo).

🚀 This is a demo pipeline. In a production scenario, you may refine the alignment, label generation, and training processes.

git clone https://github.com/your-repo/Influenza-GNN.git
cd Influenza-GNN

source venv/bin/activate   # On Windows use: venv\Scripts\activate

pip install -r requirements.txt

biopython (for sequence handling and alignment) torch and torch_geometric (for GNN implementation) pandas (for metadata processing) If torch_geometric fails to install, refer to: PyG Installation Guide

FASTA File (sequences.fasta): Contains virus sequences. Metadata CSV (metadata.csv): Should include an "Accession" column and "Collection_Date".

python main.py

✅ Load & normalize sequences ✅ Filter metadata based on available sequences ✅ Parse collection dates & select baseline sequence ✅ Perform pairwise sequence alignment (demo) ✅ Convert sequences to codon-based graphs ✅ Train a simple two-layer GNN model

Trained GNN Model (Model.pt): Can be used for inference. Printed Logs: Displays alignment results, training progress, and mutation probabilities.

• Update file paths inside main.py:

fasta_file = "sequences.fasta"
metadata_file = "metadata.csv"

• Modify GNN hyperparameters:

model = MutationGCN(in_channels=64, hidden_channels=32, num_classes=1)
optimizer = optim.Adam(model.parameters(), lr=0.01)

• Define mutation labels (dummy labels used in this demo):

graph.y = torch.zeros((num_nodes, 1), dtype=torch.float)

To load a pre-trained model (Model.pt):

from main import load_model
model = load_model("Model.pt")

We welcome contributions! To contribute:

1. Fork the repository and create a new branch:

git checkout -b feature-name

2. Make changes & commit:

git commit -m "Added feature X"

3. Push to GitHub & create a pull request.

• If you encounter any bugs, open an issue with a detailed description.

This project is licensed under the MIT License.

For questions, reach out via:

📧 Email: duashmita@gmail.com 🔗 GitHub Issues: Open an Issue Happy coding! 🚀