COID Documentation

MIRO Guidelines

Below we follow the MIRO guidelines for reporting on an ontology [1].

THIS PAGE NEEDS UPDATING

A. The basics

A.1 Combined Ontology for Inflammatory Diseases (COID) v1.0.0.
A.2 Ontology owner and creator: Samantha C Pendleton PhD student at Institute of Cancer and Genomic Sciences, University of Birmingham, UK. samanfapc@gmail.com
A.3 License, which governs the permissions surrounding the ontology (CC BY 3.0). You can use and edit COID, given you provide the appropriate credit.
A.4 Ontology under the appropriate filename coid.owl.
A.5 GitHub repository: https://github.com/sap218/coid.
A.6 Methodological framework: used Protégé software for ontology building using ICD-10 [2] and other biomedical ontologies for clinical synonyms. Synonym expansion followed the works of Pendleton et al. [3] using the TF-IDF statistical analysis to gain patient-preferred synonyms from the Patient.info forums of some inflammatory conditions.

B. Motivation

B.1 Ontology is required in research as there currently lacks a system for specifically only inflammatory disorders, symptoms, and layman-specific terms. The controlled vocabulary for these disorders are quite limited, and current biomedical layman terms are not useful for NLP tasks as layman terms should be specific to forum tasks: the need for understanding the “patient-voice”. Additionally, whilst ICD-10 is structured almost anatomtically, COID also bridges different inflammatory concepts: medical therapies, symptoms, and non-inflammatory associations.
B.2 COID is derived and it’s foundation is from ICD-10 structure. With exact synonyms from biomedical ontologies, which are currently limited in the inflammatory domain in terms of synonyms. Whilst synonyms and layman terms exist, we curate more from the patient forums.
B.3 Target audience are those in the research area looking at inflammatory conditions or the patient voice.

C. Scope, requirements, development community

C.1 The requirements of the ontology is to be an extensive vocabulary for inflammatory disorders, their associated diseases, symptoms, therapeutic interventions (medication, surgeries), investigations, complications, and more. An application ontology specific to this domain.
C.2 Development community: Samantha C Pendleton MSc, PhD student at University of Birmingham, UK. In addition to Luke T Slater PhD, Andreas Karwath PhD, and Georgios V Gkoutos PhD.
C.3 The Issues tracking system is to be used for future developements/requests/bugs/general inquiries.

D. Knowledge acquisition

D.1 Foundation built from ICD-10 codes: following the structure and then cross-reference to biomedical ontologies.
D.2 ICD-10 accessed via WHO [2] - biomedical ontologies accessed via OLS - patient forum access via Patient.info.
D.3 The included inflammatory disorders were derived from: ICD10Data & ICD LIST.

E. Ontology content

E.1 Ontology is RDF/XML OWL format, if you don’t know how to convert, create an Issue for information about alternative formats.
E.2 Specifically used Protégé to develop the ontology with Git version control.
E.3 Ontology metrics, as of v1.0.0 (see Changelog 22-05-2021) there are a total of 1185 classes, 8439 relationships and axioms, 5164 annotations, including 1763 database cross-references, and 796 layperson synonyms - curated Patient.info patient forum conversation and from PatientINF + ClinicalBERT embeddings.
E.4 Cross-references ontologies inclue: BFO, DOID, HPO, ICD-9, ICD-10, NCIT, ORDO, PATO, READCODES, SNOMED-CT, SYMP, and UBERON.
E.5 IRI for COID is: https://github.com/sap218/coid/blob/master/coid.owl#COID_00000.
E.6 Identifier generation policy: Protégé’s preferences schema is that ontology’s entities are to have the prefix COID followed by an underscore then a five digit number. All classes have COID prefix and own numeric identifier since COID avoids importing axioms from other ontologies.
E.7 Entity metadata policy: currently exact, broad, narrow, and related synonyms in the ontology are additional clinical/medical terms derived from biomedical ontologies. However layperson are the patient-preferred terms.
E.8 ICD-10 was somewhat used as an upper ontology in the sense that the structure is followed when describing the inflammatory conditions.
E.9 Currently, relationship occurs in in COID is cross-referenced to BFO. COID defined it’s own: characterised by, patient reported symptom, and symptom of.
E.10 Axiom patterns are inferred from the ontology structure itself (subclass) and by defined relationships.

F. Managing change

F.1 Sustainability plan: COID is freely available on GitHub and the creator/owner will update/change based on expert opinion or via Issue notifications. Anyone from the community can suggest new terms, but with sufficient evidence/justification - or use a pull-request. If a user makes frequent, positive changes to COID, they can be given direct access to co-maintain the ontology with the creator.
F.2 There is the Changelog to note down major changes to the ontology. In addition to Issues tracking system for entities to be removed/split/redefined.
F.3 Ontology will follow “semantic versioning” guidelines and that it will maintain as version 1 unless a major foundation change is made then it will become version 2. Otherwise additional many terms/fixes will increment to version 1.1.0, and minor/few additions/fixes will change to 1.0.1 over time (COID currently v1.0.0).

G. Quality Assurance

G.1 Successful testing on the Patient.info patient forum and ClinicalBERT embeddings. Also used Protégé’s HermiT (v1.4.3.456) reasoner to ensure a coherent and consistent ontology.
G.2 Ontology is novel and first application ontology in inflammatory disorders. Additionally this method of building (TF-IDF) was previous validated and our re-using proves a fast approach to synonym curation. COID meets the stated requirements as an ontology suitable for the biomedical domain in additon to capturing the patient voice. Moreover we show that not only is expert knowledge dervived from ontologies but also extracted from clinical letters.
G.3 An example of application is using the ontology to perform ROC AUC analysis on ClinicalBERT and PatientINF embeddings to observe how clinicians and patients discuss conditions. Additionally layman terms can be used for sentiment analysis on patient forum text, as seen from Pendleton et. al [3]. Future applications include using medical terminology x-ref codes (e.g. ICD-10) on electronic health records, or semantic similarity with COID’s implicit axiomatic structure.
G.4 Ontology not yet available on ontology repositories however freely available on GitHub.
G.5 Evidence of use: as soon as the manuscript is accepted the creator will include a reference/link here.

References

[1] Matentzoglu, N., Malone, J., Mungall, C., & Stevens, R. (2018). MIRO: guidelines for minimum information for the reporting of an ontology. Journal of Biomedical Semantics, 9(1), 6.

[2] World Health Organization (WHO). (2006). International Classification of Diseases (ICD). http://www.who.int/classifications/icd/en/

[3] Pendleton, S. C., Slater, L. T., Karwath, A., Gilbert, R. M., Davis, N., Pesudovs, K., Liu, X., Denniston, A. K., Gkoutos, G. V., & Braithwaite, T. (2021). Development and application of the ocular immune-mediated inflammatory diseases ontology enhanced with synonyms from online patient support forum conversation. Computers in Biology and Medicine, 135, 104542.