From ontology to knowledge graph with agile methods: the case of COVID-19 CODO knowledge graph

• DOI: https://doi.org/10.1108/IJWIS-03-2022-0047
• Published date: 05 October 2022
• Date: 05 October 2022
• Pages: 432-452
• Subject Matter: Information & knowledge management,Information & communications technology,Information systems,Library & information science,Information behaviour & retrieval,Metadata,Internet
• Author: Michael DeBellis,Biswanath Dutta

From ontology to knowledge

graph with agile methods:

the case of COVID-19 CODO

knowledge graph

Michael DeBellis

michaeldebellis.com, San Francisco, California, USA, and

Biswanath Dutta

Documentation Research and Training Centre, Indian Statistical Institute,

Bangalore, India

Abstract

Purpose –The purpose of this paperis to describe the CODO ontology (COviD-19 Ontology) thatcaptures

epidemiological data about the COVID-19pandemic in a knowledge graph that follows the FAIR principles.

This study took information from spreadsheets and integrated it into a knowledge graph that could be

queried withSPARQL and visualized with the Gruff tool in AllegroGraph.

Design/methodology/approach –The knowledge graph was designed with the Web Ontology

Language. Themethodology was a hybrid approach integrating the YAMO methodologyfor ontology design

and Agile methodsto define iterations and approach to requirements, testingand implementation.

Findings –The hybrid approach demonstratedthat Agile can bring the same benefits to knowledge graph

projects as it has to other projects. The two-personteam went from an ontology to a large knowledge graph

with approximately5 M triples in a few months. The authorsgathered useful real-world experience on how to

most effectivelytransform “from strings to things.”

Originality/value –This study is the only FAIR model (to the best of the authors’knowledge)to address

epidemiologydata for the COVID-19 pandemic. It also brought to light several practicalissues that generalize

to other studies wishing to go from an ontology toa large knowledge graph. This study is one of the first

studiesto documenthow the Agile approach can be used for knowledge graph development.

Keywords Agile, COVID-19, ETL, Health care, Knowledge graph, OWL, SDLC, SPARQL, Protégé,

Triplestore, FAIR

Paper type Research paper

1. Introduction

At the beginning of the COVID-19pandemic (March 2020), we began to develop an ontology

called CODO (COviD-19 Ontology) for collectionand analysis of COVID-19 data (Dutta and

DeBellis, 2020). The ontology followed the FAIR model (Wilkinson et al.,2016)for

This work is executed under the research project entitled “Integrated and Unified Data Model for

Publication and Sharing of prolonged pandemic data as FAIR Semantic Data: COVID-19 as a case

study,”funded by Indian Statistical Institute Kolkata. This work was conducted using the Protégé

resource, which is supported by grant GM10331601 from the National Institute of General Medical

Sciences of the United States National Institutes of Health. Thanks to Franz Inc. (www.allegrograph.

com) for their help with AllegroGraph and Gruff. Thanks to Dr Sivaram Arabandi, MD, for his

feedback on the CODO ontology.

IJWIS

18,5/6

432

Received1 March 2022

Revised30 June 2022

5 August2022

Accepted12 September 2022

InternationalJournal of Web

InformationSystems

Vol.18 No. 5/6, 2022

pp. 432-452

© Emerald Publishing Limited

1744-0084

DOI 10.1108/IJWIS-03-2022-0047

The current issue and full text archive of this journal is available on Emerald Insight at:

https://www.emerald.com/insight/1744-0084.htm

representing data. While otherCOVID-19 ontologies, such as CIDO, VIDO and CoVoc (more

detail provided in Section 1.2), focus on analyzing the virus at a biomolecular level, CODO

focuses on epidemiologicalissues, such as tracking how the virus was spread based on data

about social relations, geography, etc. We evolved what started as a small ontology in

Protégé to a large knowledgegraph in the AllegroGraph triplestore product from Franz Inc.

In this paper, we use theterm ontology to refer to the CODO Web Ontology Language (OWL)

model with only basic exampletest data (i.e. essentially what the ontology community refers

to as T-Box information). We use the term knowledge graph to refer to the ontology

populated with large amountsof real-world data (i.e. T-Box data with A-Box data from real

data sources).

1.1 Knowledge graphs and health care

There has been extensive work in the Semantic Web community and the health-care

domain. The Protégé ontology editor was developedin the Stanford medical school and the

original motivation for Protégé fundingwas as a tool to aid in gene research (Musen, 2015).

The major areas where knowledge graphsare used in health care are:

Drug Discovery and Repurposing –Knowledge graphs are extensively used to

analyze the molecular structure and uses of various drugs to develop new

medications and to investigate the possible repurposing of existing medications in

novel ways to treat additional diseases (Zeng et al., 2022).

Harmonization –Unlike most other domains there is already an abundance of open

vocabularies to describe medical concepts such as diseases, treatments and

symptoms. These vocabularies are used for medical research, for coding procedures

to report to insurance companies and for many other purposes. However, because of

various local and institutional requirements, a health-care organization may often

need to work with data based on different standards. For example, standards such

as SNOMED, LOINC, ICD and HL7 FHIR overlap considerably, and it is common for

large organizations to use two or more of these and other health-care standards.

Harmonization refers to the mapping of these various models to each other.

Knowledge graph technology can play a significant role in this harmonization

process. First, because of the nature of the technology with built-in support for

relations such as sameAs in OWL. Second, because most of the health-care

vocabularies are implemented as OWL models. Examples of this type of research

are Bauer et al. (2021) and Visweswaran et al. (2021).

Modeling Genetic and Biomolecular Knowledge –The rich semantics of OWL based

on Description Logic provide a powerful language to model biomolecular and

genetic knowledge. Such models can be used for artificial intelligence (AI)

applications using both deep learning and semantic AI such as rule-based systems.

Some representative examples of this work are Callahan et al. (2020) and He et al.

(2020).

Statistics –Knowledge graph technology is well suited to storing and visualizing

statistical data about various populations. This was used extensively to provide

information to the public about the COVID-19 pandemic. Some representative

examples of this work are provided in Section 1.2.

Semantic Search –One of the first applications for OWL was to facilitate a

meaningful search than a simple keyword search. For example, in a keyword

search, the keyword “OWL”will primarily return information about the bird with

the same name. One can influence the keyword search by adding additional

COVID-19

CODO

knowledge

graph

433