Disease Networks for Disease Clustering Using Graph Convolution Networks
摘要
This paper presents a technique for disease clustering using the gene ontology biological process annotation terms along with disease gene association networks with deep learning embedding. In this paper, we explore and leverage the application of Graph Convolution Networks GCN in bioinformatics; specifically in disease clustering and disease relationships. GCNs allow for aggregating two sources of information on the disease by integrating the disease network structure with disease biological process (bp) annotations from the gene ontology. The paper serves as an exploration study to examine the feasibility of integrating network structures with feature vectors for disease clustering. We obtained the disease relationship network using disease-shared genes from OMIM and DisGeNET and the disease bp features from the Gene Ontology annotation GOA database. The clustering method was performed with the embedding feature space and the experimental results show improved clustering outcomes with the feature embeddings using GCN.