ICBO 2014 Proceedings Towards the Integration of Abnormality in Diseases Yuki Yamagata, Kouji Kozaki Takeshi Imai, Kazuhiko Ohe Riichiro Mizoguchi Department of Knowledge Science Department of Medical Informatics Research Center for Service Science I.S.I.R, Osaka University Graduate School of Medicine Japan Advanced Institute of Science 8-1 Mihogaoka, Ibaraki, Osaka, Ja- The University of Tokyo and Technology pan 7-3-1, Hongo, Bunkyo-Ku, Tokyo, 1-1 Asahidai, Nomi, Ishikawa, Ja- yamagata@ei.sanken.osaka-u.ac.jp Japan pan Abstract—Knowledge of abnormalities is important for un- from our abnormality ontology to other biomedical concepts, derstanding diseases. A number of resources provide terms relat- we can establish mutually complementary relationships be- ed to abnormalities in biomedicine. In this paper, we investigate tween biomedical concepts in different levels, which will con- the differences in the hierarchical structure of these biomedical tribute to ensuring interoperability across biomedical resources. resources and discuss issues of reuse and integration with regard Our goal is to provide not only a theory for a better understand- to abnormal states based on ontological theory. Then, we show a ing of abnormal states but also useful information for clinical solution for integrating them by linking abnormal states in our practice. We are planning to integrate abnormalities in the def- abnormality ontology to other resources according to the mean- inition of diseases in the Department of Cardiovascular Medi- ing of the concepts at each level. cine and other medical departments at the University of Tokyo Keywords—ontology; abnormality; disease Hospital in our ontology with external biomedical resources at each level of meaning as a concrete example. I. INTRODUCTION This paper is organized as follows. In Section 2, we intro- In order to understand diseases, one must first capture the duce our ontology of abnormalities. In Section 3, we examine abnormal states in diseases adequately. They are observed as the characteristics of biomedical resources, discuss the problem symptoms by patients or as signs by clinicians in clinical find- with them based on ontological theory and propose a solution ings. Clinical test data can also provide evidence for the exist- for integration with respect to abnormal states. Finally, in Sec- ence of abnormal states. In addition, in basic research, through tion 4, we present concluding remarks and a give an outline of analysis of disease models of animals, many researchers make future work. efforts to understand how causative agents are related in the etiological process. Moreover, abundant knowledge about ab- II. ABNORMALITY ONTOLOGY normalities is available in scientific articles. The above obser- This section provides an overview of our abnormality on- vation shows that abnormality is a key factor for capturing tology. diseases, assisting in the interoperability between basic re- search and clinical medicine for integrating a wide variety of A. Three-Layer Ontological Model of Abnormal States knowledge across domains. We have been involved in the de- In order to develop an is-a hierarchical tree of abnormali- velopment of a disease ontology [1]. As part of this research, ties, it is important to conceptualize them from a consistent we have focused on abnormal states in the definition of diseas- viewpoint. To this end, we have been developing an abnormali- es and have rigorously systematized an abnormality ontology ty ontology [2, 3] having a three-layer structure: [2, 3]. • Level 1: Generic abnormal states A number of terminologies and standard vocabularies have been developed for many years, and recently, ontologies have • Level 2: Object-dependent abnormal states also been constructed in the biomedical domain. They offer • Level 3: Specific disease-dependent abnormal states useful data, and some of these terms include abnormality con- cepts. In order to make efficient use of these resources, we The top-level categories define very basic and generic con- need a solution that ensures interoperability between abnormal- cepts, for example, "small in area," "hypofunction," etc., which ity knowledge across domains. To achieve this, first, one has to are commonly used not only in clinical medicine but also in elucidate one's own perspectives of the resources before using other domains. Level 2 concepts are dependent on objects. In them and to make the meanings of concepts explicit. the lower level of the tree, concepts are designed to represent abnormalities at specific human organ / tissue / cell levels. For In this paper, we discuss differences of existing biomedical example, by specializing "small in area" at Level 1, " narrowed resources based on ontological engineering theory with respect cross-sectional area of tube ", where the cross-sectional area of to abnormalities. Next, we investigate the relationships be- a tubular structure has become narrowed, is defined at Level 2, tween abnormal states in our abnormality ontology and corre- and this is further specialized in the definitions "vascular steno- sponding terms in biomedical resources. By mapping concepts sis" (blood vessel-dependent), "arterial stenosis", "coronary 7 ICBO 2014 Proceedings artery stenosis" (coronary artery-dependent), and so on. Level 3 form. For example, in terms of the state of hypertension 3, our concepts are captured as specific disease-dependent (context- model ensures interoperability among the forms , , and stenosis" at level 2 is defined as a constituent of ischemic heart . Therefore, our model realizes interopera- disease at Level 3. In our ontological approach, common con- bility between test data and abnormal states in the definition of cepts can be kept distinct from specific ones and can be appro- diseases. In practice, we found that more complicated terms, priately defined according to their context. such as adding modifier words (e.g., transient hypertension), and three types of compound words (e.g., “blood pressure” as As of 11 May 2013, our ontology has 21,669 abnormal an OA type: < blood (O), pressure (A)>). We have developed a states constituted of 6,302 diseases across 13 medical depart- guideline and deal with them as a variation of data representa- ments, and among them, clinicians have currently refined con- tion. cepts of 9,985 abnormal states constituted of 1,602 diseases across five major departments. III. UTILIZATION OF BIOMEDICAL RESOURCES B. Representation of Abnormal State In this section, we investigate the hierarchy of "coronary ar- In medicine, abnormal states are interpreted from the di- tery stenosis" in existing biomedical resources and perform a verse perspectives of specialists such as clinicians, pathologists, comparison between our ontology and existing biomedical biologists, geneticists, and so on, and correspondingly a variety resources to make them interoperable with each other with of representations of abnormal states are used. Therefore, we respect to abnormalities. classified the abnormal states into three categories: a property 1 A. Charestictics of biomedical resources (e.g., hypertension), a qualitative representation (e.g., blood 1) SNOMED-CT and MeSH Terminologies pressure is high), and a quantitative representation (e.g., blood pressure 180 mmHg). In previous work, we proposed a Proper- a) SNOMED-CT: Systematized Nomenclature of Med- ty-Attribute interoperable representation framework for ab- icine-Clinical Terms (SNOMED-CT) is a clinical terminology normal states [2, 3] on the basis YAMATO [4]. first developed by the College of American Pathologists (CAPs) and is currently maintained by the International Health We captured all abnormal states as properties2 represented Terminology Standards Development Organization (IHTSDO) by a tuple: , e.g., . We specified the property by decomposing it into a tu- lection of clinical terms that contain more than 310,000 con- ple: . The Attribute Value can be either a Qualitative Value (Vql) or a Quantitative Value cepts. It is widely used as an international standard vocabulary. (Vqt). For example, "arterial stenosis" is decomposed into SNOMED-CT has a hierarchical structure. The root concept of as a qualitative repre- the hierarchy is named "SNOMED-CT," and there are 19 top- sentation, or as a quan- level categories, including Clinical finding/disorder, Body titative representation. Then, we introduce "Object" to identify structure, Organism, Substance, and other things important for the target object, and we represent an abnormal state as a triple: clinical health. Most of the abnormal states are included in the . This is the category Clinical finding/disorder. In the hierarchical tree, basic form in our representation model of abnormalities. In concepts are linked by is-a relationships. SNOMED-CT al- addition, we introduce "Sub-Object (SO)" as an advanced rep- lows multiple inheritances. In addition, one concept can have resentation for what will be focused on. For example, in the relationships other than is-a, like "finding site," "method," case of "hyperglycemia", since the glucose concentration (A) "clinical course" and so on, to connect concepts between dif- means the ratio of the focused object (SO) relative to the whole ferent categories, e.g., "occlusion of artery has finding site mixture (O), the representation of "hyperglycemia" is a quad- arterial structure." ruple, . In another case of an advanced representation, "Colonic poly- b) MeSH: Medical Subject Headings (MeSH) is a the- posis" is described as . Our model can deal with both clinical test data and Medicine and is used for indexing biomedical articles [6]. The abnormal states in the definition of diseases. The clinical test 2014 MeSH contains 27,149 terms as descriptors (MeSH data can be represented in the form (OAVqt), which can be converted chical tree called "MeSH tree" is organized into 16 categories, into a property representation form (OPVp) via a qualitative representation Most of the concepts of abnormal states are classified into the diseases category. MeSH also allows multiple inheritances. 2) PATO and HPO Phenotype Ontologies: In biomedical domains, scientists observe entities through experiments or 1 The property discussed here is based on ontological theory, not Web Ontol- clinical findings to capture abnormal states. Therefore, repre- ogy Language (OWL) property which means a link (relation) between two nodes. 2 3 A state derived from or associated with a property (P) is defined as: "a Our model can describe the mechanism “hypertension” more specificity by temporal entity derived by a time-indexed property (P) in which the bearer using other factors; e.g., accumulation of atheroma in arterial wall →decreased elasticity of pant. arterial wall →hypertension. 8 ICBO 2014 Proceedings sentations of knowledge about abnormalities are usually given results in multiple inheritance from two parents in the PATO as phenotypes. hierarchical tree: one is abnormal states, and the other is prop- Phenotypic Quality Ontology (PATO) is an ontology of erty (quality). For example, "decreased area" in PATO has two phenotypic qualities for annotating biological phenotypes super-classes, "decreased size" and "area" (Fig. 1). Such mul- across species [7]. PATO provides qualities to describe pheno- tiple inheritance makes things complicated, and it is difficult typic information, such as size, color, weight and so on, and to understand the underlying meaning of concepts. If comput- qualities have abnormal states as subclasses of each quality; for ers integrate PATO concepts and other resources in a naïve example, size has lower concepts like "increased/decreased way, inappropriate consequences might be derived unexpect- length," "increased/decreased volume," "dwarf-like", in addi- edly. For example, HPO provides the concept of "coronary tion to length, area, and volume. PATO is widely used as a artery stenosis," and an incorrect relation such as "coronary standard vocabulary for biological measurement in bio- stenosis is-a area" might be derived by a naïve combination of communities. PATO also plays a role as a reference ontology PATO and HPO. "Area" is not an abnormal state. In our on- for databases of specific species, such as Drosophila [8]. tology, parameters such as pressure, area, concentration, and The Human Phenotype Ontology (HPO) is an ontology for so on, are properly dealt with as an attribute (A), so that there human specific phenotypes [9]. It has been developed using is no possibility of deriving inappropriate consequences such phenotypic information from the human genetic diseases data- as the one shown above. base Online Mendelian Inheritance in Man (OMIM) [10] and medical articles, and provides over 10,000 terms. HPO classifies human pheno- types into three categories: 1) mode of inher- itance, 2) onset and clinical course, and 3) phe- notypic abnormalities. Abnormal state concepts are mainly subclasses of 3) phenotypic abnor- malities. 3) LOINC: Logical Observation Identifiers, Name and Codes (LOINC) was developed by the Regenstrief Institute as a universal code for ex- changing clinical test data [11]. It is widely used in 185 countries. LOINC provides six fields (component (analyte), property, timing of the measurement, sample type, scale type, and method). By combining the contents of each field with colons, the name of a clinical test can be described. For example, "a test for glucose Fig. 1 Is-a hierarchy of coronary artery stenosis. tolerance about after 2 hours serum glucose for With respect to the intrinsic nature of abnormal states, 100g oral" is represented by "GLUCOSE^2H POST 100 G states can change, and their essential characteristic is under- GLUCOSE PO:MCNC:PT:SER/PLAS:QN." stood with respect to how the corresponding attribute (A) has B. Ontological Issues and Solution with Regard to the value (V), which can change as time goes. Then, one can Abnormality find that the authentic "is-a" relation should be "decreased 1) PATO and HPO: Both PATO and HPO representations area is-a decreased size" rather than "decreased area is-a area", are based on property representations in accordance with the and our model adopts the former. upper ontology Basic Formal Ontology (BFO) [12]. Therefore, Next, we discuss ontological issues in HPO. The HPO hi- one could say that these representations are similar to our erarchy is based primarily on where the abnormal states occur property representation of abnormal states. However, in PATO (e.g., arterial stenosis is-a abnormalities of peripheral arteries). and HPO, the relationship between Property (P) and Attribute Therefore, it is difficult to know the is-a relation of "stenosis" (A) & Value (V) is unspecified, which leads to a lack of in- comprehensively. In the HPO tree, ad hoc creation of correct teroperability with clinical test data. The data values are most- is-a relationships is found, such as "renal artery stenosis is-a ly quantitative, and clinical test data are represented in terms arterial stenosis." To make matters worse, HPO has no generic of A and V. To achieve interoperability with clinical test data, concepts at the upper level nor any information about what not only the OP form but also the OAV form is required. Our attribute takes what value, and we cannot capture the essential model can deal with both representation forms and can property of the state as "stenosis." Furthermore, specialization achieve interoperability between OP and OAV, which will be using the part-whole relation appears in the is-a hierarchy, e.g., of great assistance in medical practice , which is misleading. (P) from Attribute (A) in the same way as BFO does, which 9 ICBO 2014 Proceedings With respect to abnormal states, it is important to capture 2) SNOMED-CT and MeSH: The development of the identity of the state by revealing "where and how its intrin- SNOMED-CT and MeSH started before ontological engineer- sic attribute takes a particular abnormal value." On the basis of ing had become mature, and thus they have some problems such understanding, our ontology adopts a single is-a relation- ontologically [14]. SNOMED-CT and MeSH do not have any ship to inherit the intrinsic nature of abnormal states. For ex- formal upper ontology. This allows for multiple inheritance ample, all lower classes of "small in area" inherit the property that results in complicated relations. Such a structure may lead of "the area (A) is small (V)," such as" narrowed cross- to inconsistencies in daily clinical practice, since the organiza- sectional area of tube," "vascular stenosis," and so on. tion of concepts is not principled, and hence relationships In practice, however, only a single hierarchical structure among concepts lack consistency. As shown in Fig. 1, in seems incompatible with various perspectives in biomedicine. SNOMED and MeSH, abnormal states and diseases are mixed In order to support different perspectives, we developed a up in the hierarchy, as is the case in HPO. In MeSH, the upper technology for dynamically generating an on-demand classifi- class "coronary stenosis" is a disease, namely, "coronary dis- cation hierarchy [13]. For instance, if researchers want to ease," and at an even higher level, diseases are defined, like know the classification of abnormal states in terms of anatom- "heart disease" and "cardiovascular disease." In SNOMED, ical structure, by using the is-a relation of Object (O) for each the upper concept of "coronary artery stenosis" is a "coronary abnormal state, a subclassification according to anatomical occlusion," and furthermore, at even higher levels, there are structure abnormalities can be generated (Fig. 2). The parton- various concepts, such as "heart disease," "vascular (blood omy of organs can also be used for generating a hierarchy, vessel) finding," "disorder of cardiovascular system," "disor- which is similar to the part-of-whole relationship of anatomi- der of soft tissue" and so on. cal structure: e.g., "atrium abnormality" is classified as a sub- SNOMED-CT terms have been used for electronic health category of "cardiac abnormality." record (EHR) systems. Because SNOMED-CT does not dif- ferentiate abnormal states from diseases, serious clinical prob- lems may occur. For example, imagine a case where a clini- cian examines an angiographic image and, by using the SNOMED-CT term "coronary artery stenosis," records it for evidence of "abnormal states" in the EHR system. Due to in- heriting properties from one of the upper concepts in SNOMED-CT, namely, "disease," there is a possibility of deriving an erroneous consequence that the "abnormal state" is a "disease." Our ontology makes a distinction between diseas- es, disease-dependent abnormal states, and disease- independent abnormal states. Therefore, there is no possibility Fig. 2 Reorganization of is-a relationships of abnormal states in of deriving inappropriate consequences, which is important for terms of anatomical structure. computer processing. In order to manage high-level clinical Another critical issue with HPO is that HPO does not dif- knowledge in EHR, we need a reliable method for representa- ferentiate abnormal states from diseases and organize them in tion, and our ontological model provides us with high reliabil- one hierarchy, which may be misleading and confusing in ity and sophisticated technology for realizing interoperability clinical practice. In HPO, the parent of "coronary stenosis" is between heterogeneous pieces of clinical information. considered to be not an abnormal state but a disease, namely, 3) LOINC: LOINC was excluded from the above- "coronary disease." However, as shown in Fig. 1, "coronary mentioned comparative research, because LOINC does not disease" has three parents: "arterial stenosis," "coronary artery deal with concepts related to abnormal states. abnormality", and "atherosclerosis," which seem to be abnor- LOINC provides the form O (So) A and is useful for in- mal states. Therefore, there is no guarantee that a computer teroperability among various clinical test data. However, can derive the correct answer to the question of whether "cor- LOINC does not have Value (V). To realize interoperability onary artery stenosis" is a disease or an abnormal state. between clinical test data and abnormal states, a Quantitative Our ontology enables us to make a distinction between ge- Value (Vqt) is needed in the representation form. Our model neric abnormal states, object-dependent ones, and disease- can deal with quantitative data in the OAV form, and, there- specific ones with a unified representation and allows us to fore, we can transform it into the OP form of abnormal states. specialize concepts to the required granularity with consisten- As a result, our model has the ability to maintain interoperabil- cy. We conceptualize a disease as an entity represented in ity between clinical test data and abnormal states in diseases. terms of abnormal states, and we deal with abnormal states Some readers might think, "Why not reuse other resources and diseases as different entities. Therefore, we can focus on such as clinical terminology or existing ontologies?", because the intrinsic nature of the states themselves from one view- it seems that by combining existing resources, many abnormal point and can develop an ontology from the viewpoint of state, states would be easily covered. Such good candidates for reuse without mixing up the viewpoint of disease. concerning abnormal states are PATO, HPO, LOINC, and 10 ICBO 2014 Proceedings SNOMED-CT. As we illustrated in the problem above, just Mapping our ontology to other resources for integrating reutilizing existing resources cannot integrate all knowledge various data related to abnormalities will bring benefits to the about abnormal states. It is too difficult to find all inappropri- users of other resources, too. First, one can find concepts from ate usages of is-a and resulting misclassifications from the generic to specialized terms easily by referring to the single is- a tree in our abnormality ontology. For example, although huge and heterogeneous system of concepts and to make an HPO does not care about consistent is-a relationships in terms effort to modify them. Furthermore, since each resource has its of "stenosis", by referring to "arterial stenosis" at Level 2 in our own viewpoint, integrating them into a unified perspective ontology through mapping, HPO users can get the is-a relation- must be a hard task because it necessarily requires compara- ships: "arterial stenosis is-a vascular stenosis is-a narrowed tive analysis and validation of accuracy in integrated concepts. cross-sectional area of tube is-a small in area." Since "small Unified theoretical considerations have resulted in interop- area" is linked to a PATO concept, via our ontology, users erability between various representation forms, which will might find orthologous concepts of other species. Specifically, enable us to establish a computer-understandable model for human phenotypes can be linked to the phenotypes of model abnormal states. We need more sophisticated organization of organisms, e.g., mouse, rat, etc., if the set composed of Attrib- related representations, including quantitative and qualitative ute (A) and Value (V) are identical, and the Object (O) has structural similarity. PATO2YAMATO aims to integrate phe- data and knowledge at higher levels of abstraction about ab- notype descriptions residing in differently structured compari- normal states in the definition of diseases to exploit all of them son contexts [16]. By applying PATO2YAMATO, mapping of in a consistent manner. Our model is the first one to make concepts across species and integrating knowledge from vari- such exploitation possible, and will be of great assistance in ous species may be possible. medical practicee. The differences of resources are summarized in Table 1. TABLE I. COMPARISON OF BIOMEDICAL RESOURCES C. Integration of biomedical abnormal states Fig. 3. Integration of abnormal states in biomedicine. 1) Integration of abnormal states 2) Integration of components of abnormal states As illustrated in Section 2, our ontology provides three lev- els of abnormal states from generic to disease-specific. We are also planning to link the components of the representation of abnormal Level 1 in our ontology defines generic concepts corre- states to other resources. We will try to connect the Object (O) sponding to the PATO concepts, and our Level 1 concepts can to concepts in FMA (The Foundational Model of Anatomy be mapped to related PATO concepts (Fig. 3). The lower Level ontology) [17]. 2 concepts are human anatomical structure-dependent abnor- mal states, which correspond to the HPO concepts. By creating 3) Integration of biomedical articles links between Level 2 concepts and HPO concepts, it will be By mapping MeSH terms, it will be possible to retrieve bi- possible to navigate from the HPO concepts to the upper gener- omedical articles related to abnormal states or diseases. If we ic concepts of PATO. Level 3 provides disease-specific ab- obtain new findings of the constituents of diseases, we can add normal states, such as "myocardial ischemia in ischemic heart new relationships to the causal chains of the diseases, which disease," "chest pain in angina pectoris", and so on. In the re- might contribute to the elucidation of the etiological mecha- vised version 11 of the International Classification of Diseases nism. Our mapping is also useful for MeSH users to understand (ICD), diseases contain information of "causal properties" [15], how their research subjects are involved in various abnormal and therefore, we are planning to map our Level 3 concepts to states in the human body in diseases. This will contribute to the the corresponding concepts in the ICD. Level 3 abnormal states progress of biomedical research. are described in the causal chains of diseases [1]. By mapping our disease concepts of disease ontology to the ICD, ICD users D. Concluding Remarks can understand the causal relationships of the abnormal states A large volume of data and concepts related to abnormal in diseases. Our ontology can also allow users to navigate re- states is currently available in existing resources. However, lated concepts in other resources, such as HPO, PATO, etc. there are no resources that cover all levels of abnormal states 11 ICBO 2014 Proceedings from generic to disease-specific. We performed a comparison Linked Data (LD)[18], and have made available a browsing between our ontology and existing resources, and identified the system that links our data to DBPedia [19] and 3D images of issue that the heterogeneous levels of meanings in the different related anatomical parts provided by BodyParts3D [20]. Next, resources and multiple perspectives prevented us from reusing we are planning to map our abnormal states to other resources and integrating them. This motivated us to develop an abnor- using LD and to scale up to applications requiring more com- mality ontology from the generic level to the disease-specific plicated knowledge. level. ACKNOWLEDGMENT Our medical ontology project started seven years ago. Since then, it has been refined and revised several times by discus- A part of this research is supported by the Japan Society for sion with both ontologists and clinicians. Our ontology will the Promotion of Science (JSPS) through its “FIRST Pro- play a role in proper guidelines for giving an ontological point gram”" and the Ministry of Health, Labour and Welfare, Japan. of view in various controlled vocabularies, and will lead to the The authors are deeply grateful to Drs. Ryota Sakurai, Natsuko development of a consistent hierarchical structure with a uni- Ohtomo, Aki Hayashi, Takayoshi Matsumura, Satomi Terada, fied representation. Mapping our ontology with other resources Kayo Waki, and other, at The University of Tokyo Hospital for at each level of meanings will contribute to ensuring interoper- describing disease ontology and assisting us with their broad ability across biomedical resources. Since our approach allows clinical knowledge. We also would like to thank other team users to navigate from generic concepts to specific concepts in members, Drs. Yoshimasa Kawazoe, Masayuki Kajino, and other domains by following links, it offers complementary Emiko Shinohara from The University of Tokyo for useful information. We are currently applying the concepts of abnor- discussions related to biomedicine. mal states in the definition of diseases in the Department of REFERENCES Cardiovascular Medicine and several other departments at The University of Tokyo Hospital in our ontology and mapping [1] R. Mizoguchi, K Kozaki, H Kou, Y Yamagata, T Imai, K Waki, K Ohe, “River flow model of diseases,” in ICBO2011, 2011, pp. 63-70. them to external biomedical resources, and this work is ex- [2] Y Yamagata, K Kou, K Kozaki, T Imai, K Ohe , R Mizoguchi, pected to be completed in the near future. Level 1 generic con- “Ontological model of abnormal states and its application in the medical cepts have previously been developed by ontologists, and Lev- domain,” in ICBO2013, 2013, pp. 28-33. el 3 disease-specific concepts were also described. Currently, [3] Y Yamagata, K Kozaki, T Imai, K Ohe , R Mizoguchi, “An ontological we are developing and enriching Level 2 concepts to link each modeling approach for abnormal states and its application in the medical Level 3 concept to upper-level common concepts. By develop- domain.,” J Biomed Semantics, 2, 5:23, 2014. ing Level 2, we will be able to find more commonalities across [4] R. Mizoguchi, “YAMATO: Yet another more advanced top-level diseases. For example, in cardiovascular medicine, "increased ontology.” In Proceedings of the Sixth AOW 2010, 2010, pp:1-16. blood creatine kinase (CK) concentration in acute myocardial [5] SNOMED-CT [http://www.ihtsdo.org/snomed-ct/]. infarction" is defined by a clinician at Level 3. Next, "increased [6] MeSH [http://www.nlm.nih.gov/mesh/meshhome.html] blood CK" at Level 2 is defined and mapped to "elevated se- [7] GV Gkoutos, EC Green, AM Mallon A Blake, S Greenaway, JM rum creatine phosphokinase" in HPO. Then, the generic con- Hancock, D Davidson, “Ontologies for the description of mouse phenotypes,” Comp Funct Genomics, vol.5, pp: 545-51, 2004. cept "increased concentration" is defined at Level 1 in our on- [8] G Grumbling, V Strelets, “FlyBase: anatomical data, images and tology, which is mapped to "increased concentration" in PATO. queries,” Nucleic Acids Res, vol. 34, pp: D484-488, 2006. We can find commonalities with other concepts in the defini- [9] S Köhler, SC Doelken, CJ Mungall, et al, “The Human Phenotype tion of diseases in other medical departments in our ontology. Ontology project: linking molecular biology and disease through For example, "increased blood CK concentration in muscular phenotype data,” NAR, vol. 42 (Database issue), pp.D966-74. 2014. dystrophy" used in the Neurology Department has commonali- [10] OMIM [http://www.ncbi.nlm.nih.gov/omim] ty with "increased blood CK concentration," and moreover, [11] LOINC [http://loinc.org/] "increased blood cholesterol concentration in hyperlipidemia" [12] Grenon P, Smith B, Goldberg L, “Biodynamic ontology: applying BFO has commonality with the same generic concept, "increased in the biomedical domain,” Stud Health Technol Inform, vol. 102, pp. concentration." We are planning to map the disease "acute 20-38, 2004. myocardial infarction" in our disease ontology to "acute myo- [13] K Kozaki et al., “Dynamic Is-a Hierarchy Generation System Based on cardial infarction" in the ICD. In our ontology, diseases are User's Viewpoint,” in procceing of JIST, LNCS 7185, pp.96-111, 2012. defined as causal relationships of abnormal states, and from [14] S Schulz, B Suntisrivaraporn, F Baader, “SNOMED CT's problem list: this mapping, ICD users will be able to know the causal rela- Ontologists' and Logicians' therapy suggestions,” Stud Health Technol Inform. vol.129(Pt 1), pp:802-806, 2007;. tionship, "decreased blood flow" causes "myocardial ische- [15] ICD 11 [I http://www.who.int/classifications/icd/revision/en/] mia," which results in "myocardial necrosis" in acute myocar- [16] H Masuya, R Mizoguchi: “An advanced strategy for integration of dial infarction. This causal relationship is also available to biological measurement data ,” in ICBO2011, 2011, pp. 79-86. users of HPO by following the linked concept of abnormal [17] C Rosse, JVL Mejino. “A reference ontology for biomedical states. With respect to "myocardial ischemia," by mapping this informatics: the Foundational Model of Anatomy,” J Biomed Inform, to the MeSH term "myocardial ischemia," we can collect arti- vol. 36, pp: 478-500., 2003. cles that are related to the concept. [18] K Kozaki, Y Yamagata, T Imai, K, et al., “Publishing a disease ontologies as linked data,”. in Proc. of JIST2013, 2014, pp. 110-128． In this way, our approach allows users to navigate various [19] DB pedia [http://dbpedia.org/] abnormality knowledge across domains, which will create a [20] BodyParts3D, The Database Center for Life Science bridge between basic research and clinical medicine. We pub- [http://lifesciencedb.jp/bp3d/] lished the causal chains in our disease ontology in the form of 12