Building a spatially-explicit knowledge graph integrating health, pathogen and environmental data for COVID-19 analysis

Abstract:

Understanding COVID-19 infection rates and factors requires integration of information from multiple disciplines. While many diverse pandemic-related datasets have been made available online over a short time by hundreds of projects around the world, they are often published with little coordination or reliance on common standards and vocabularies. We present a community effort to address this challenge by organizing such diverse information fragments into a COVID-19 knowledge graph (KG). The KG integrates knowledge assertions derived from datasets and text sources, in particular those describing virus strains and other pathogen information, population health statistics, social-economic and demographic factors and community interaction patterns affecting disease rates and transmission, and environmental data. It is designed to help researchers answer questions about differences in outbreak characteristics across locations with different socio-demographic or environmental conditions, compare population health effects of viral strains, or trace different patterns of imported vs community infection spread. We present initial experience creating and populating the KG, in particular as related to organizing and disambiguating location information found in many COVID-19 sources. For standardizing and indexing location information we use a combination of text analysis and disambiguation tools developed in the EarthCube’s Data Discovery Studio and in other open source and commercial projects. We discuss issues of integrating geospatial information in a KG as implemented in a Neo4J environment, compare location disambiguation results from different tools, and demonstrate several experimental spatial analysis dashboards created to explore, visualize and query KG information online. The knowledge graph, along with code for populating and querying the graph organized as a collection of Jupyter notebooks, is available online (https://github.com/covid-19-net/covid-19-community) and is open to community contributions.