Graph Learning Beyond Connectivity: A hands-on introduction to GNNs and graph foundation models

Biological systems are naturally relational: genes, proteins, cells, and organisms interact through networks whose edges carry both direction and meaning. This beginner-friendly seminar introduces graph learning by moving from graph basics to the central idea of message passing in graph neural networks (GNNs), then to the intuition behind graph foundation models that can be adapted across tasks. Using synthetic graph data, we will examine why activation versus inhibition and source-to-target direction should be represented explicitly rather than collapsed into ordinary connectivity. A short segment will show how simple topological summaries can reveal connected structure, cycles, and potential feedback. We will use a single Google Colab notebook that runs in a web browser and connects each concept to compact, interpretable code.