Toward a Coronavirus Knowledge Graph

Area of Science:

• Biomedical Informatics
• Computational Biology
• Data Science

Background:

• The COVID-19 pandemic necessitates integrated knowledge for drug discovery and research.
• Existing datasets are fragmented, hindering comprehensive analysis of the viral domain.

Purpose of the Study:

• To construct a unified coronavirus knowledge graph (KG) by merging diverse data sources.
• To enhance the understanding of COVID-19 related biological entities and their relationships.

Main Methods:

• Integrated the Analytical Graph (AG) drug discovery dataset with the CORD-19 article collection.
• Performed entity disambiguation on CORD-19 data using Wikidata.
• Populated the KG with over 21,700 genes, 2,500 diseases, and 94,000 phenotypes, defining 27 relationship types.

Main Results:

• The resulting KG contains a rich network of chemo genomic and biological entities relevant to COVID-19.
• Analysis of an ego-centered network around angiotensin-converting enzyme (ACE) provided COVID-19 insights.
• Identified significant COVID-19-related paths between entities like hydroxychloroquine and IL-6 receptor.

Conclusions:

• The developed KG offers a valuable resource for COVID-19 research and drug discovery.
• The KG's structure facilitates the exploration of complex biological relationships and potential therapeutic strategies.