Subgrouping Rare Disease Patients Leveraging the Human Phenotype Ontology Embeddings

In the USA, rare diseases are defined as those affecting fewer than 200,000 patients at any given time. It usually takes substantial time and long journey for rare disease patients to seek care before receiving a correct diagnosis. Making the right phenotypic characterization is the initial step to speed up such differential diagnosis at early time and the Human Phenotype Ontology (HPO) is a comprehensive knowledgebase supporting this goal. Previously, we have constructed various node embeddings for the HPO incorporating heterogeneous biomedical knowledge repositories. In this study, we applied unsupervised learning strategies over different HPO embeddings, aiming to further subgroup rare disease patients based on phenotypic characterizations.