TY - GEN
T1 - Constructing Node Embeddings for Human Phenotype Ontology to Assist Phenotypic Similarity Measurement
AU - Shen, Feichen
AU - Liu, Sijia
AU - Wang, Yanshan
AU - Wang, Liwei
AU - Wen, Andrew
AU - Limper, Andrew H.
AU - Liu, Hongfang
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/7/16
Y1 - 2018/7/16
N2 - The Human Phenotype Ontology (HPO) was developed to be a semantically computable vocabulary that captures the phenotypic abnormalities found in human diseases discovered through biomedical research. Usage of this ontology facilitates the translation between genotype and phenotype. Many studies have been conducted to accelerate the implementation of precision medicine into clinical practice by utilizing the informative contents provided in the HPO. No work, however, has been done in constructing a distributed representation for nodes in HPO to provide a deep insight of phenotypic similarities by analyzing its graph structure. Node2vec is a model for generating node embeddings based on large networks. In this study, we constructed node embeddings for the HPO leveraging node2vec to assist phenotypic similarity measurement. A downstream application on link prediction driven by HPO embedding achieved 0.81 ROAUC and 0.75 F-measure. A use case study was conducted on idiopathic pulmonary fibrosis (IPF) and we demonstrated the potential possibility of using HPO embeddings in assisting phenotypic similarity measurement.
AB - The Human Phenotype Ontology (HPO) was developed to be a semantically computable vocabulary that captures the phenotypic abnormalities found in human diseases discovered through biomedical research. Usage of this ontology facilitates the translation between genotype and phenotype. Many studies have been conducted to accelerate the implementation of precision medicine into clinical practice by utilizing the informative contents provided in the HPO. No work, however, has been done in constructing a distributed representation for nodes in HPO to provide a deep insight of phenotypic similarities by analyzing its graph structure. Node2vec is a model for generating node embeddings based on large networks. In this study, we constructed node embeddings for the HPO leveraging node2vec to assist phenotypic similarity measurement. A downstream application on link prediction driven by HPO embedding achieved 0.81 ROAUC and 0.75 F-measure. A use case study was conducted on idiopathic pulmonary fibrosis (IPF) and we demonstrated the potential possibility of using HPO embeddings in assisting phenotypic similarity measurement.
KW - Human Phenotype Ontology
KW - Node embeddings
KW - Phenotypic similarity
UR - http://www.scopus.com/inward/record.url?scp=85051003522&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85051003522&partnerID=8YFLogxK
U2 - 10.1109/ICHI-W.2018.00011
DO - 10.1109/ICHI-W.2018.00011
M3 - Conference contribution
AN - SCOPUS:85051003522
T3 - Proceedings - 2018 IEEE International Conference on Healthcare Informatics Workshops, ICHI-W 2018
SP - 29
EP - 33
BT - Proceedings - 2018 IEEE International Conference on Healthcare Informatics Workshops, ICHI-W 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 6th IEEE International Conference on Healthcare Informatics Workshops, ICHI-W 2018
Y2 - 4 June 2018 through 7 June 2018
ER -