Graph Neural Networks for Social and Biological Network Analysis

(National AI at Cedars-Sinai)

This project explores how Graph Neural Networks (GNNs) can model and understand complex, interconnected systems in both social and biological domains. Its central goal is to teach and demonstrate how GNNs capture relationships and patterns in non-Euclidean data like social interactions, biological cell networks, and medical imaging where traditional machine learning methods fall short.

Through this project, participants build, train, and evaluate deep learning models such as Graph Convolutional Networks (GCNs), GraphSAGE, and Graph Attention Networks (GATs) to analyze datasets ranging from physician collaboration networks and lymphocyte detection in tissue slides to fake news propagation on Twitter and brain connectivity patterns in neuroscience.

The broader goal is to advance the understanding of how graph-based AI can be applied in healthcare and biomedical research, helping detect, predict, and explain relationships between entities- whether they’re doctors, cells, or neural regions. By combining computational modeling with real-world health and social data, the project contributes to Cedars-Sinai’s mission of bridging AI innovation with medicine and public health systems.