Graph Neural Networks for Chemistry - Day 1

Explore the fundamentals of Graph Neural Networks (GNNs) in chemistry through this comprehensive lecture from the 2024 Machine Learning for Drug Discovery Summer School at Mila. Delve into the anatomy of graphs and their application to molecular structures, understanding atoms and bonds as key components. Learn about molecular property prediction and the intricacies of designing effective GNNs, including concepts like WL expressivity, attention mechanisms, and pretraining techniques. Discover important datasets used in the field and examine scaling strategies for larger models. Investigate the potential of multimodality in enhancing GNN performance for chemical applications. Conclude with a Q&A session to clarify concepts and discuss future directions in this rapidly evolving field.

Introduction
Overview
Graph Anatomy
What is a molecule
Atoms and bonds
Molecular property prediction
Designing a GNN
WL Expressivity
Attention
Pretraining
Data Sets
Scaling
Multimodality
Questions