Ontology Representation for Cholangiocarcinoma Anuwat Pengput 1,2, and Alexander D. Diehl 1 1 Department of Biomedical Informatics, Jacobs School of Medicine and Biomedical Science, University at Buffalo, Buffalo, NY, 14203, USA 2 Huaimek District Public Health Office, Ministry of Public Health, Kalasin, 46170, Thailand Abstract Introduction: Cholangiocarcinoma is a critical public health problem in Thailand. Several research projects have been conducted and data related to CCA have been collected to solve this problem. Data about CCA are found in varied sources such as research-based databases and electronic health records that have been collected and stored using different methods and standards. The objective of this study is to develop the Cholangiocarcinoma Ontology (CCAO) to describe findings related to cholangiocarcinoma in a structured and standardized way in order to integrate and analyze data from these diverse sources. Methods: CCAO has been developed based on data collection forms (CCA forms) of the Cholangiocarcinoma Screening and Care Program (CASCAP). The forms contain data elements about demographics, ultrasound findings, confirmatory diagnoses, final staging diagnoses, and post-operative and follow-up outcomes. These data elements were used to search the Ontobee web browser for matching ontology classes in existing ontologies. Ontology classes from various sources were extracted using a ROBOT tool and imported to CCAO, and new CCAO classes for unmatched classes were added to CCAO manually. CCAO is an application ontology beneath the Basic Formal Ontology (BFO) along with the Ontology of General Medical Science (OGMS), the Information Artifact Ontology (IAO), and the Ontology for Biomedical Investigations (OBI). Results: Based on the CCA forms we developed 210 novel CCAO classes and created 108 CCAO classes based on NCI Thesaurus classes. We reused classes from various domain ontologies including the Phenotype And Trait Ontology (PATO), the Ontology of Biological Attributes (OBA), the Cell Ontology (CL), the Ontology of Medically Related Social Entities (OMRSE), and Drug Ontology (DRON). Imported classes in CCAO were reorganized under the top-level classes such as OGMS:‘clinical finding’, OGMS:‘disorder’, and OBI:‘conclusion based on data’. Moreover, we generated logical definitions for many CCAO classes. Conclusion: CCAO is reusable, interoperable, and easily integrated with related datasets, as well as being human and machine readable. It is compatible with future expansion to represent relevant evidence and knowledge that is not be part of this initial version. CCAO is publicly available on Github (https://github.com/Buffalo-Ontology-Group/CCA-Ontology). Keywords 1 cholangiocarcinoma, biomedical ontology, basic formal ontology 1. Introduction fish are the key factors for liver fluke, Opisthorchis viverrini (O. viverrini), infections. O. viverrini infections produce hepatic bile ducts Cholangiocarcinoma (CCA) is a major and portal connective tissue inflammation. problem in Southeast Asia (SEA). The prevalence Chronic infections and inflammation have been of CCA in SEA is much higher than other areas in indicated to be risk factors for the development of the world. Culture and traditions of eating raw, multiple stages of carcinogenesis [1, 2]. fermented, pickled, and undercooked cyprinid ICBO 2022, September 25-28, 2022, Ann Arbor, MI, USA EMAIL: anuwatpe@buffalo.edu (A. 1); addiehl@buffalo.edu (A. 2) ORCID: 0000-0002-0273-1531 (A. 1); 0000-0001-9990-8331 (A. 2) ©️ 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR Workshop Proceedings (CEUR-WS.org) CEUR ceur-ws.org Workshop ISSN 1613-0073 Proceedings In Thailand, CCA is a common malignancy. information [14]. The objective of this study is to Several policies have been deployed over the last create an application ontology, CCAO to describe 40 years to prevent CCA, but the survival rates of in a structured and standardized way various types CCA patients are still poor [3]. The Chol- of information and findings related to CCA. angiocarcinoma Screening and Care Program (CASCAP), established by Khon Kaen Univer- sity, Thailand in 2015, aims to eliminate O. 2. Methods viverrini infections and CCA [4]. CASCAP is a prospective cohort study CCAO is being developed based on data items including screening and patient cohorts. Data is of the CASCAP forms which are used to collect collected based on six separate data collection data about demographics, ultrasound findings, forms: CCA-01, “Demographic Information diagnoses and treatments, and post-operative Enrollment,” CCA-02, “Ultrasound,” CCA-02.1, follow-up outcomes from targeted populations in “Confirmatory Diagnosis,” which is used to area of Thailand where OV and CCA are endemic confirm suspected CCA participants from [4, 5]. All variables names and data elements from ultrasound screenings in CCA-02 based on CT scan, MRI, or other procedures, CCA-03, the CASCAP forms were used to search on the Ontobee web browser for ontology classes [15] “Diagnosis and Treatment,” CCA-04, “Final for matching with existing ontologies. Staging Diagnosis,” and CCA-05, “Post Operation Follow-up” [4, 5]. Moreover, electronic health records (EHR) from general and 2.1. Development of CCAO community hospitals in Thailand also contain data and information of patients with, or suspected of We have developed CCAO based on the suffering from CCA including symptoms, clinical CASCAP forms 1, 2, 2.1, 3, 4, and 5. The CCA findings, treatments, and diagnoses [6]. forms were translated from Thai to English. We These databases represent data elements in evaluated the English versions of the CASCAP different ways. EHR uses International Statistical forms to ensure the correct translation. After the Classification of Disease and Related Health English language forms were prepared, we Problems, 10th revision, Thai Modification (ICD- mapped all data items to classes and classes in 10-TM), which is being used as the Thai standard existing ontologies using Ontobee [15] and then for morbidity and mortality coding in health evaluated the quality of the mapped data items services statistics, as well as for billing and with existing ontology classes. payment [7, 8], while the CASCAP study We have imported classes into CCAO from represents data and information as research-based various different sources including the Uberon data elements that capture details about CCA multi-species anatomy ontology (Uberon), the more specific than those in ICD-10-TM. Thus, it Phenotype And Trait Ontology (PATO), the is a challenge to work with data from different Ontology of Biological Attributes (OBA), the Cell sources and standards. Ontology (CL), the Ontology of Medically In order to integrate data from different Related Social Entities (OMRSE) [16] and Drug sources, the Cholangiocarcinoma Ontology Ontology (DRON) [16-21]. We used ROBOT to (CCAO) is being built as an application ontology extract classes from external ontologies and under the Basic Formal Ontology (BFO) and the generate import files (.owl) [22], and applied the BFO-compatible ontologies including the Syntactic Locality Module Extractor (SLME) Ontology of General Medical Science (OGMS), method to extract classes using the BOT (The the Information Artifact Ontology (IAO), and the BOT, or BOTTOM) algorithm. The resulting Ontology for Biomedical Investigations (OBI) [9- ontology module contains mainly the classes in 12]. CCAO relies upon BFO to provide an upper- the seed, plus all their super-classes and the inter- level framework to structure the ontology. relations between them. All import files along Ontologies have long been used to designate with the upper-level ontologies including BFO, all entities within an area of reality and all OGMS, IAO, and OBI, were imported directly to relationships between those entities in a way that Protégé for creating CCAO [9, 11, 12]. make them interpretable by both humans and The National Cancer Institute Thesaurus computers [13], and the application of ontologies (NCIT) [23] provides comprehensive information in medical and scientific research is a response to related to CCA. However, we chose not to import need to reuse the voluminous and complex classes from NCIT directly, because of the 2 difficulty in merging the NCIT classes into the retained only classes related to CCA in order to BFO-OGMS hierarchy. As a result, we based keep CCAO small and precise. All imported many classes in CCAO on similar NCIT classes classes were placed under upper-level ontology and have included references to those classes in classes from BFO, OGMS, IAO, and OBI. CCAO. We took advantage of the information content in NCIT class definitions in building 3. Results CCAO as an OBO Foundry compliant ontology. For example, in CCAO, the ‘cholangiocarcinoma’ 3.1. Summary of ontology classes class is defined as a “A adenocarcinoma that arises from a bile duct.”. The class is assigned to CCAO includes upper-level ontology classes a CCAO_ID and references the original ‘NCIT: from BFO, OGMS, OBI, and IAO. We developed Cholangiocarcinoma’ class URI using the 210 new CCAO classes based on data items in the skos:closeMatch annotation property of the CCA forms. We created 117 CCAO classes based Simple Knowledge Organization System in order on NCIT classes as well as one class based on a to indicate the similarity in meaning to the Mammalian Phenotype Ontology (MP) class. external class [24]. [27]. Finally, we reused classes from various We found many data items in the CCA forms domain ontologies (Table 1) including 13 classes that did not map to existing ontology classes, so from PATO such as ‘abnormal’, ‘calcified’, we created new classes for these data elements. ‘edematous’, ‘mucoid’, and ‘morphology’; 8 We did a literature review for each new ontology classes from OBA such as ‘hepatic vein class to define its meaning based on principles of morphology’, ‘hepatic portal vein morphology’, best practice in classes, definition, and and ‘lymph node morphology’; 2 classes from classification with desiderata for controlled CL, ‘neoplastic cell’ and ‘malignant cell’; 2 medical vocabularies to improve face value of classes from OMRSE, ‘admission process’ and ontology classes [9, 25]. ‘patient discharge’; and 1 class from DRON, For instance, we created a new class, ‘praziquantel oral tablet’. ‘intrahepatic bile duct mass-forming cholangiocarcinoma’, which is defined as “An Table 1 intrahepatic cholangiocarcinoma of the Summary of classes in CCAO intrahepatic bile duct that has a mass-forming Ontologies Number tumor morphology, consisting of a single solid CCAO classes based on CCA forms 210 and lobulated mass with no connection CCAO based on NCIT classes 108 macroscopically discernible with a bile duct and CCAO based on MP class 1 characterized by irregular but well-defined and not encapsulated borders,” to represent the Uberon 23 intrahepatic bile duct mass-forming data item on PATO 13 CCA-04 form. We also asserted its parent to be OBA 8 ‘intrahepatic cholangiocarcinoma’ and created a CL 2 logical definition as follows: ‘intrahepatic OMRSE 2 cholangiocarcinoma’ and (hasQuality some DRON 1 ‘mass-forming tumor morphology’). The classes unique to CCAO were assigned CCAO identifier numbers (CCAO_ID). Each new 3.2. Mapping of NCIT classes class was added manually using Protégé with a unique IRI in the form of an OBO Foundry We used 108 NCIT classes as the basis of new persistent URL (PURL) [26]; for instance, a CCAO classes. These classes are needed to periductal fibrosis class is assigned to represent data elements on the CCA02-CCA05 http://purl.obolibrary.org/obo/CCAO_00141. forms and are classified under the top level These IRIs do not currently resolve, but we ontology classes including: OGMS:‘clinical intend to apply for admission to the OBO Foundry finding’ such as ‘Bismuth-Corlette perihilar in the near term and have thus chosen to use a cholangiocarcinoma classification’, ‘cancer TNM compatible IRI format. finding’, ‘intrahepatic bile duct cancer TNM After importing the external ontology classes, finding v8’, and related classes; OGMS:‘disorder’ there were a number of irrelevant classes included such as ‘fibrosis’, ‘cirrhosis’, ‘ascites’, in CCAO. We removed irrelevant classes and ‘neoplasm’, and related classes; OGMS: 3 ‘diagnostic process’ such as ‘biopsy’, ‘computed tumor’, and ‘mixed type tumor morphology’. tomography’, ‘diagnostic ultrasound’, and related Additionally, we developed new CCAO classes classes; OGMS:‘therapeutic procedure’ such as for other CCA tumor morphologies such as ‘cancer therapeutic procedure’, ‘percutaneous ‘cholangiocarcinoma-encased hepatic artery’, and trans-hepatic biliary drainage’, ‘bypass’, ‘surgical ‘cholangiocarcinoma-positive lymph node along procedure’, ‘biliary stenting’, and related classes; hepatoduodenal ligament’. and BFO:‘process’ such as ‘activity’, ‘referral,’ In CCA-03 form “Diagnosis and Treatment,” and ‘withdraw’. We used one MP class ‘dilated most variables and data elements about diagnostic bile duct’ as the basis of a CCAO class, and process, treatment, and complications in this form placed it under OGMS:disorder, in order to could be mapped to NCIT classes. We generated represent this as a disorder rather than a new CCAO classes based on these NCIT classes phenotype. for supporting the CCA-03 form, such as ‘extended right hepatectomy’, ‘surgical resection 3.3. Creation of new CCAO classes of hilar cholangiocarcinoma’, ‘exploratory lap- arotomy of liver including biopsy,’ and ‘palliative percutaneous transhepatic biliary drainage’. A number of variables and data elements in the The CCA-04 form is used to collect data about CCA forms do not match to existing ontology the results of pathological diagnoses, which are classes. We created 210 new CCAO classes along final staging diagnoses. We developed new with new definitions based on data dictionary of CCAO classes and definitions to classify types of the CCA forms and scientific literature. The CCA based on this form and review of the participants’ self-reported variables and data literature. In Figure 1, CCA was categorized by a elements in the CCA-01 form were used as the tumor site in bile duct including intrahepatic, basis of new CCAO classes modeled as subtypes perihilar, and distal CCA along with mass- OBI:‘conclusion based on data’, for instance, forming, intraductal, and periductal infiltrating ‘conclusion about participant report about history tumor morphology. The CCA types were also of fecal examination for liver fluke egg’, classified by the histology and mucinous type. ‘conclusion about participant report about consumption of raw fresh-water fish or raw • intrahepatic cholangiocarcinoma fermented fish’, ‘conclusion about participant =def. - A cholangiocarcinoma found in report about history of treatment with antiparasitic any site of the intrahepatic biliary tree that drug’, and ‘conclusion about participant report arises from the intrahepatic bile duct about having relatives with cholangiocarcinoma’. epithelium The CCA-02 form variables and data elements Logical definition - cholangiocarcinoma are about ultrasound screening. We created new and (overlaps some 'intrahepatic bile duct'). CCAO classes and developed new definitions along with appropriate parent classes as needed. • perihilar cholangiocarcinoma There classes were classified under top ontologies =def. - A cholangiocarcinoma found in classes including: OGMS:‘clinical finding’ such the common hepatic duct between the second- as ‘suspected cholangiocarcinoma’, ‘finding of order biliary ducts (the left and right hepatic thickening of wall of gallbladder’, and ‘finding ducts) and the cystic duct insertion. about kidney parenchyma with atypical abnormal Logical definition - cholangiocarcinoma function’; OGMS:‘image finding’ such as and (overlaps some 'common hepatic duct'). ‘hepatic mass ultrasound echo finding’, left lobe hepatic mass high echo finding’, ‘left lobe hepatic • distal cholangiocarcinoma mass low echo finding’, and ‘left lobe hepatic =def. - A cholangiocarcinoma found in mass mixed echo finding’; and OGMS:‘disorder’ the common bile duct between the cystic duct such as ‘periductal fibrosis’ and subtypes, and and the ampulla of Vater (except Klatskin ‘hepatic calcification’; and OGMS:‘diagnostic tumors and ampulla of Vater cancer), which process’ including ‘liver diagnostic ultrasound’, includes mid common bile duct tumors and ‘hepatic parenchymal ECHO’. (between the junction with the cystic duct and In CCA-02.1 form “Confirmatory Diagnosis,” the junction with the pancreas) and distal we developed new CCAO classes for CCA tumor (intrapancreatic) bile duct tumors. morphology including ‘mass-forming’, Logical definition - cholangiocarcinoma ‘periductal infiltrating’, ‘intraductal intrahepatic and (overlaps some 'common bile duct'). 4 ‘distal periductal infiltrating cholangiocarcinoma’ is defined as “A distal cholangiocarcinoma that has a periductal infiltrating tumor morphology in which the tumor spreads along the biliary tree, without mass formation,” and has been given the logical definition 'distal cholangiocarcinoma' and (hasQuality some 'mass-forming tumor morphology')) (illustrated in Figure 2). Similarly, ‘distal intraductal cholangiocarcinoma’ has the logical definition 'distal cholangiocarcinoma' and (hasQuality some 'intraductal intrahepatic tumor morphology'). 4. Discussion CCAO is designed to cover all data items in the CCA forms using a BFO-based hierarchy and relying on the principles of OBO Foundry in order Figure 1: Cholangiocarcinoma hierarchy in CCAO to avoid repetition of efforts and to facilitate reuse of and compatibility with domain ontologies [28]. We also created an extension of cancer TNM We made attempts to find and reuse existing staging (T: primary tumor, N: regional lymph domain ontology classes related to the CCA nodes, and M: distant metastasis) including new forms. We were able to match many data elements classes such as ‘intrahepatic bile duct cancer pt4a with the existing classes, and we created new TNM finding v8’, ‘intrahepatic bile duct cancer classes when no existing ontology classes could pt4b TNM finding v8,’ ‘perihilar bile duct cancer be found and because some data elements were pt3a TNM finding v8’, and ‘perihilar bile duct very specific in the domain of CCA in Thailand, cancer pt3b TNM finding v8’. Moreover, new such as the consumption of raw fish dishes CCAO classes were created to represent the (cyprinoid fish). metastatic malignant neoplasm in lung, Schuler and Ceusters [29] reported that diaphragm, and lung or pleura. On the other hand, building application ontologies appeared to be a we did not create any new class for the CCA-05 challenging job, and described a number of form “Post Operation and Follow Up,” because problems they encountered. In line with their we were able to reuse existing ontologies to work, the main problem we experienced was to represent variables and data elements. find adequate ontologies and adequate classes within them. We found some relevant ontologies to representing items on the CCA forms, such as NCIT, were not BFO-compatible and rarely followed the principles of the OBO Foundry. Although many data elements could be matched with existing ontologies classes in our first attempt at mapping, we then decided not to reuse these classes because they were placed in inconsistent hierarchies or had suble differences in definition that did not match our needs for CCAO. NCIT is of particular interest in that it is an extraordinary source of cancer knowledge and vocabulary; unfortunately, it is not a BFO- Figure 2: Logical definition of distal periductal compatible ontology. In our early efforts, we used infiltrating cholangiocarcinoma in Protégé NCIT as a core domain ontology and extracted relevant classes using ROBOT for inclusion in CCAO. However, the extracted NCIT module Furthermore, we provided 41 new CCAO contained many classes irrelevant to CCA and the classes with logical definitions. For example, 5 overall hierarchy was incompatible with BFO and 5. Conclusions OGMS. NCIT contains top-level concept classes such as ‘Conceptual Entity’, ‘Disease, Disorder, CCAO has been developed to represent data or Finding’, and ‘Drug, Food, Chemical or about CCA using best practices in ontology Biomedical Material’, that proved impossible to development. The ontology is publicly available classify under the upper level ontologies used in at Github (https://github.com/Buffalo-Ontology- CCAO. Group/CCA-Ontology) and is compatible with Eventually, we removed NCIT classes and future expansion to represent new evidence and created similar CCAO classes to use in our knowledge not be part of this initial version. ontology. This also allowed to use Uberon for all anatomical terms rather than NCIT anatomical classes. During the process of mapping the data 6. Acknowledgements elements to Uberon, we found that there was no class ‘wall of gallbladder’ available in Uberon, Part of the research reported in this publication although this term exists in the Foundational was supported by the Royal Thai Government Model of Anatomy Ontology (FMA). In order to Scholarship, Praboromarajchanok Institute of limit the mixing of hierarchies as much as Heath Workforce Development, Ministry of possible, we submitted a request to Uberon editors Public Health, Thailand. We would like to to add a new class, ‘wall of gallbladder’, which acknowledge Professor Werner Ceusters, MD for was added to Uberon and reused in CCAO. evaluating CCAO and providing valuable CCAO has 41 logical definitions in this initial suggestions, and we would like to thank Assistant version, which we will improve upon in the future. Professor Dr. Kavin Thinkhamrop, Faculty of We are working in a parallel fashion to develop a Public Health, Khon Kaen University, Thailand first order logic axiomatization using Common for providing useful information related to Logic Interchange Format (CLIF) in order to CASCAP and CCA forms. render the CCAO compatible with BFO2020 axiomatization [30] and to allow for more 7. References complex reasoning. CLIF can work with time indexing and negation. We will use CLIF to create axiomatization for all classes in CCAO and then [1] Hughes T, O'Connor T, Techasen A, Namwat generate an owl-compatible version based on this N, Loilome W, Andrews RH, et al. work. This approach will be used to verify the Opisthorchiasis and cholangiocarcinoma in consistency and satisfiability of CCAO. Southeast Asia: an unresolved problem. Int J Gen In upcoming work, CCAO will be used to Med. 2017;10:227-37. analyze associated patient datasets from the [2] Sripa B, Kaewkes S, Sithithaworn P, Mairiang Kalasin Provincial Public Health, Ministry of E, Laha T, Smout M, et al. Liver fluke induces Public Health, Thailand that include verbal and cholangiocarcinoma. PLoS Med. 2007;4:e201. ultrasound screening data and EHR-recorded [3] Kamsa-Ard S, Luvira V, Suwanrungruang K, symptoms and diagnoses related to CCA. The Kamsa-Ard S, Luvira V, Santong C, et al. criteria for participants include people who live in Cholangiocarcinoma Trends, Incidence, and Kalasin, Thailand and who participated and Relative Survival in Khon Kaen, Thailand From provided the information in all stages of verbal 1989 Through 2013: A Population-Based Cancer screening, ultrasound screening, and made Registry Study. J Epidemiol. 2019;29:197-204. hospital visits related to CCA. Ontology-based [4] Khuntikeo N, Chamadol N, Yongvanit P, enrichment analysis and predictive models will be Loilome W, Namwat N, Sithithaworn P, et al. performed in order to understand the complexity Cohort profile: cholangiocarcinoma screening of risk factors for CCA. and care program (CASCAP). BMC Cancer. This project has been approved by University 2015;15:459. at Buffalo Institutional Review Board (IRBID: [5] Cholangiocarcinoma Foundation of Thailand. STUDY00006059). Isan Cohort. Khon Kaen University, Thailand: CASCAP: Cholangiocarcinoma and Care Program; 2016. [6] Cholangiocarcinoma Foundation of Thailand. Summary of the situation of cholangiocarcinoma, a review of patient charts. Khon Kaen, Thailand: 6 Ministry of Public Health and Khon Kaen base and resources. Nucleic Acids Research. University; 2016. 2019;47:D1018-D27. [7] Schulz S, Rodrigues JM, Rector A, Spackman [20] Vasilevsky CMRSNMuk- K, Campbell J, Ustün B, et al. What's in a class? pJBMBpDKdCTHMHN. obophenotype/bio- Lessons learnt from the ICD - SNOMED CT attribute-ontology. Zenodo; 2022. harmonisation. Stud Health Technol Inform. [21] Diehl AD, Meehan TF, Bradford YM, Brush 2014;205:1038-42. MH, Dahdul WM, Dougall DS, et al. The Cell [8] Bureau of Policy and Strategy. International Ontology 2016: enhanced content, Statistical Classification of Diseases and Related modularization, and ontology interoperability. Health Problems 10th Revision Thai Journal of Biomedical Semantics. 2016;7. Modification. Nonthaburi, Thailand: Ministry of [22] Jackson RC, Balhoff JP, Douglass E, Harris Public Health; 2016. NL, Mungall CJ, Overton JA. ROBOT: A Tool [9] Arp R, Smith B, Spear AD. Building for Automating Ontology Workflows. BMC Ontologies with Basic Formal Ontology: The Bioinformatics. 2019;20:407. MIT Press; 2015. [23] National Cancer Institute. NCI Thesaurus [10] Bandrowski A, Brinkman R, Brochhausen (NCIt). National Institutes of Health, National M, Brush MH, Bug B, Chibucos MC, et al. The Cancer Institute, USA: The OBO Foundry; 2020. Ontology for Biomedical Investigations. PLoS [24] Alistair Miles, Bechhofer S. SKOS One. 2016;11:e0154556. Reference. SKOS Simple Knowledge [11] Ceusters W. An information artifact ontology Organization System perspective on data collections and associated Reference2009. representational artifacts. MIE2012. p. 68-72. [25] Cimino JJ. Desiderata for controlled medical [12] Scheuermann RH, Ceusters W, Smith B. vocabularies in the twenty-first century. Methods Toward an ontological treatment of disease and Inf Med. 1998;37:394-403. diagnosis. Summit Transl Bioinform. [26] Jackson R, Matentzoglu N, Overton JA, Vita 2009;2009:116-20. R, Balhoff JP, Buttigieg PL, et al. OBO Foundry [13] Ashburner M, Ball CA, Blake JA, Botstein in 2021: operationalizing open data principles to D, Butler H, Cherry JM, et al. Gene Ontology: evaluate ontologies. Database. 2021;2021. tool for the unification of biology. Nature [27] Smith CL, Eppig JT. The mammalian Genetics. 2000;25:25-9. phenotype ontology: enabling robust annotation [14] Yu AC. Methods in biomedical ontology. and comparative analysis. Wiley Interdiscip Rev Journal of Biomedical Informatics. 2006;39:252- Syst Biol Med. 2009;1:390-9. 66. [28] Irshad Ally WC. Challenges in Realism- [15] Xiang ZM, C; Ruttenberg, A; He, Y. Based Ontology Design: A Case Study on Ontobee: A Linked Data Server and Browser for Creating an Ontology for Motivational Learning Ontology Terms. Proceedings of the 2nd Theories. International Conference on International Conference on Biomedical Biomedical Ontology (ICBO). Bolzano, Italy: Ontologies (ICBO). Buffalo, NY, USA: CEUR CEUR Workshop Proceedings (CEUR-WS.org); Workshop Proceedings (CEUR-WS.org); 2011. p. 2021. 279-81. [29] Schuler JC, Ceusters WM. The Problems of [16] Hicks A, Hanna J, Welch D, Brochhausen M, Realism-Based Ontology Design: a Case Study in Hogan WR. The ontology of medically related Creating Definitions for an Application Ontology social entities: recent developments. J Biomed for Diabetes Camps. AMIA Annu Symp Proc. Semantics. 2016;7:47. 2017;2017:1517-26. [17] Hanna J, Joseph E, Brochhausen M, Hogan [30] Buffalo Developers Group. Basic Formal WR. Building a drug ontology based on RxNorm Ontology (BFO) 2020 Common Logic. Github; and other sources. J Biomed Semantics. 2021. 2013;4:44. [18] Mungall CJ, Torniai C, Gkoutos GV, Lewis SE, Haendel MA. Uberon, an integrative multi- species anatomy ontology. Genome Biology. 2012;13:R5. [19] Köhler S, Carmody L, Vasilevsky N, Julius O, Danis D, Gourdine J-P, et al. Expansion of the Human Phenotype Ontology (HPO) knowledge 7