Graph Neural Networks for Identifying Complex Gene–Disease Associations in Rare Syndromes
Graph Neural Networks for Identifying Complex Gene–Disease Associations in Rare Syndromes
Ali Alizadeh1
1) Department of Computer Engineering, Iran University of Science and Technology, Tehran, Iran
IOI Code
:
AAAB-ABCEZF
Publication :
2nd International Congress on Engineering Sciences and Sustainable Urban Development(dtuconf.com)
Abstract :
Rare syndromes often have very few known cases, making it difficult to identify new gene–disease relationships. We propose a GNN-based framework to detect complex gene–disease associations by leveraging heterogeneous biomedical graphs. Key data sources include OMIM and Orphanet for rare diseases, DisGeNET for gene–disease associations, and HGNC for gene nomenclature. We construct a graph with gene nodes, disease nodes, and edges representing known associations and biological interactions. We experiment with multiple GNN architectures (graph convolutional networks, GraphSAGE, graph attention networks) trained for link prediction. Our evaluation uses held-out associations and metrics like ROC-AUC and PR-AUC. Results demonstrate that GNN models significantly outperform baseline methods. For example, our GCN model achieved ~0.95 AUC on DisGeNET data.
Keywords : graph neural network gene disease association rare disease knowledge graph