Graph learning has become one of the most successful approaches for analysing complex biomedical data such as brain connectivity networks, molecular interactions, and patient similarity graphs. However, most existing graph neural networks are developed for individual diseases or specific datasets, limiting their ability to generalise across different clinical applications.
Inspired by the success of foundation models in natural language processing and computer vision, this project seeks to develop general-purpose graph foundation models capable of learning transferable representations from large-scale medical data. Rather than training separate models for each disease, the proposed framework will investigate self-supervised and multimodal graph learning approaches that can be adapted to multiple downstream tasks with minimal labelled data.
The research will contribute toward the development of reusable graph representations for next-generation medical AI systems and digital health applications.
Required knowledge
- Python programming
- Machine learning and deep learning
- Graph Neural Networks (GNNs) or graph representation learning
- Self-supervised and transfer learning
- Mathematical foundations (linear algebra, probability, optimisation)
- Basic knowledge of biomedical data analysis (desirable)
