Graph Representation Learning and Its Applications to Biomedicine

Explore graph representation learning and its applications in biomedicine through this insightful lecture. Delve into SubGNN, a subgraph neural network for learning disentangled subgraph embeddings that capture complex topology, including structure, neighborhood, and position within a graph. Discover how these methods have been applied to predict disease treatments, verified through laboratory experiments, and to identify safer drug combinations with fewer side effects. Learn about the development of actionable representations that allow for meaningful interpretation of model predictions. Cover topics such as machine learning on graphs, graph neural networks, subgraph challenges, attention mechanisms, and hypergraphs. Gain valuable insights into the intersection of graph theory, machine learning, and biomedical applications.

Introduction
Presentation
Machine learning on graphs
Graph neural networks
Agenda
Problem formulation
Topology
Subgraphs
Why are subgraphs challenging
Subgroup Neural Network
Recap
Results
Audience Question
Summary
Motivation
Core problem
Key insight
Input data
Task
Wrapup
Resources
Datasets
Attention mechanism
Hypergraphs