Supervision: Ali Hariri

Project type: Semester project (master) Master thesis

Available

Project Description: Graph Neural Networks (GNNs) have emerged as powerful tools for modeling and analyzing graph-structured data. However, many biological networks, such as protein interaction networks, gene regulatory networks, and metabolic pathways, are characterized by long-range dependencies that standard GNNs struggle to capture effectively.

This semester project focuses on exploring and implementing methods to incorporate long-range dependencies in GNNs for biological network applications. The project will investigate techniques such as virtual nodes, hierarchical clustering, and rewiring graph structures. The goal is to apply these techniques to real-world biological datasets to improve performance on tasks like node classification, link prediction, or graph-level classification, while gaining insights into biological processes.

Key Objectives:

Study existing approaches for handling long-range dependencies in GNNs. Apply these methods to biological networks, leveraging domain-specific characteristics. Analyze performance improvements and evaluate the biological interpretability of the results.

Requirements:

Strong programming skills and experience with PyTorch. Familiarity with Graph Neural Networks (experience with PyTorch Geometric is a plus). Interest in computational biology or willingness to learn about biological networks.

Deliverables

An implemented GNN framework that accounts for long-range dependencies in biological networks. A report or presentation summarizing findings, methods, and results. Optionally, a contribution to an open-source repository or publication draft if the results are significant. This project provides an excellent opportunity to work at the intersection of graph learning and biological applications, with the potential to uncover novel insights into complex biological systems.

Supervisor: Ali Hariri, a PhD student working on long-range interactions with Graph Neural Networks. For more information, please reach out to me at ali.hariri@epfl.ch.