Graph Neural Networks in Life Sciences: Opportunities and Solutions

Abstract

Graphs (or networks) are ubiquitous representation in life sciences and medicine, from molecular interactions maps, signaling transduction pathways, to graphs of scientific knowledge and patient- disease-intervention relationships derived from population studies and/or real-world data, such as electronic health records and insurance claims. Recent advance in graph machine learning (ML) approaches such as graph neural networks (GNNs) has transformed a diverse set of problems relying on biomedical networks that traditionally depend on descriptive topological data analyses. Small- and macro- molecules that were not modeled as graphs also saw a bloom in GNN-based algorithms improving the state-of-the-art performance for learning their properties. Comparing to graph ML applications from other domains, life sciences offer many unique problems and nuances ranging from graph construction to graph- level, and bi-graph-level supervision tasks. The objective of this tutorial is twofold. First, it will provide a comprehensive overview of the types of biomedical graphs/networks, the underlying biological and medical problems, and the applications of graph ML algorithms for solving those problems. Second, it will showcase four concrete GNN solutions in life sciences with hands-on experience for the attendees. These hands-on sessions will cover: 1) training and fine-tuning GNN models for small-molecule property prediction on atomic graphs, 2) macro-molecule property and function prediction on residue graphs, 3) bi-graph based bind- ing affinity prediction for protein-ligand pairs, and 4) organizing and generating new knowledge for drug discovery and repurposing with knowledge graphs. This tutorial will also instruct the attendees to develop in two extensions of the software library Deep Graph Library (DGL), including DGL-lifesci and DGL-KE, so that they could jumpstart their own graph ML journey to advance life science research and development.