Hyperbolic embeddings improve disease diagnosis from patient graphs
Hyperbolic Graph Representation Learning for Differential Diagnosis on Biomedical Knowledge Graphs
Artificial IntelligenceMachine Learning
Summary
Medical data can be complicated because it connects diseases, genes, symptoms, and patients in many ways. The authors studied a way to represent this data using a mathematical space called hyperbolic space, which naturally fits hierarchical relationships like ontologies. They found that this method worked well even when mixing different types of medical information and helped rank diseases likely affecting a patient. This approach could make computer tools for diagnosing genetic diseases more accurate and efficient.
What this means in practice
- •For hospital data teams: Rank likely genetic diseases for patients by combining ontology and patient data using hyperbolic graph embeddings.
- •For bioinformatics software developers: Build diagnostic tools that use low-dimensional hyperbolic embeddings to efficiently capture complex biomedical hierarchies and relationships.
Authors
Pietro Miotto, Lucia Mellini, Tommaso Marzi, Cesare Alippi, Elena Casiraghi, Alberto Paccanaro, Giorgio Valentini, Mauricio Soto-Gomez
Abstract
Biomedical knowledge graphs combine ontology-derived hierarchies with transversal associations among heterogeneous entities such as phenotypes, diseases, genes, proteins, and patients. This hybrid structure raises the question of whether hyperbolic embeddings, which naturally capture tree-like organization, remain useful beyond purely hierarchical graphs. We present a preliminary study of hyperbolic graph representation learning for Mendelian-disease differential diagnosis on a patient-integrated biomedical graph. Experiments on isolated ontology subgraphs show that hyperbolic models achieve strong performance in substantially lower dimensions than Euclidean baselines. We then evaluate the models on a link-prediction task that ranks candidate diseases for each patient. Results suggest that hyperbolic embeddings can exploit biomedical hierarchical structure while supporting diagnostic reasoning over heterogeneous patient-level graphs.