1. The Relevance of Ontologies in Biology and Medicine Stefan Schulz Freiburg University Hospital Medical Language and Ontology Group (MediLOG) Department of Medical Informatics

2. 448 in the last year

3. www.symbiomatics.org 1. Medical Genetics Databases and Initiatives 2. Gene Expression Information in Medical Diagnostics & Prognostics 3. Modelling & Simulation of Biological Structures & Processes/Diseases 4. Data Integration from Biosensors & Med. Devices with clinical information systems 5. Integration of patient molecular data in Electronic Health Records 6. Systems for Clinical Decision Making 7. Semantic Interoperability and Ontologies in Biomedicine 8. Technologies for Biomedical Information Integration 9. Data Interoperability & Standards 10. Connecting Biobanks to large scale databases to enable data mining 11. Patient Risk Profiling and Lifestyle Management 12. Applied Pharmaceutical Research 13. Clinical and Ethical Issues related to biomedical data processing Identification of the “Thirteen most highly prioritised areas” : White Paper, June 2006

4. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

5. A cruise through the archipelago of ontologies GO ChEBI MA FBcv FMA WordNet ICD GALEN SNOMED FAO GENIA NCI MeSH TA BRENDA GROCL MedRa

6. MeSH: Medical Subject Headings MeSH Medical Subject Headings

11. MeSH: Medical Subject Headings GO Gene Ontology

15. MeSH: Medical Subject Headings ICD International Classification of Diseases

19. MeSH: Medical Subject Headings Word NET

22. MeSH: Medical Subject Headings SNOMED Clinical Terms

24. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

25. ONTOLOGY: Unresolved Terminological Confusion… Artifacts for ordering domain entities, relating word meanings or providing semantic reference  Vocabularies  Terminologies  Thesauri  Concept Systems  Classifications  Ontologies

26. ONTOLOGY: Unresolved Terminological Confusion… Artifacts for ordering domain entities, relating word meanings or providing semantic reference  Vocabularies  Terminologies  Thesauri  Concept Systems  Classifications  Ontologies

27. ONTOLOGY: Unresolved Terminological Confusion… Artifacts for ordering domain entities, relating word meanings or providing semantic reference  Vocabularies  Terminologies  Thesauri  Concept Systems  Classifications  Ontologies

28. Different scientific traditions: Biology, Medicine, Philosophy, Logic, Linguistics, Library and Information Science, Computer Science, Cognitive Science Different philosophical schools of thinking: Platonism, Aristotelian Realism, Conceptualism, Relativism, Idealism, Postmodernism, Constructivism, Nominalism, Tropism,… ONTOLOGY: Unresolved Terminological Confusion…

29. Ontologies / Terminological Systems come in different flavors ExtractionOfForeignBodyFromStomachByIncision ≡ RemovalOfForeignBodyFromDigestiveSystem ∏ RemovalOfForeignBodyFromStomach ∏ IncisionOfStomach ∏  has-part.(  Method.RemovalAction ∏  DirectMorphology.ForeignBody) ∏  has-part.(  Method.IncisionAction ∏  ProcedureSite.stomachStructure) Nodes and Links(In)formal Definitions domain or region of DNA [GENIA]: A substructure of DNA molecule which is supposed to have a particular function, such as a gene, e.g., c-jun gene, promoter region, Sp1 site, CA repeat. This class also includes a base sequence that has a particular function.

30. What do the nodes in Ontologies / Terminological Systems stand for? concepts classes entities categories terms sets synsets universals sorts properties types descriptors descriptors names

31. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

32. Purposes of Ontologies: General Semantic Interoperability Terminology control Knowledge extraction Knowledge management Natural Language Processing Document retrieval Formal reasoning about knowledge structures

33. Purposes of Ontologies: Medicine Support of clinical coding (diagnoses, procedures): Accounting Health Statistics Support of Biomedical Science: Interoperability between heterogeneous databases Indexing of biomedical literature

34. Purposes of Ontologies: Biology Data and information retrieval and analysis Semantic Annotation of Genes, Proteins in terms of localization, pathways, functions.. Intelligent text mining of literature abstracts: "Bibliomics"

35. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

36. Mapping the space of Ontology instead of providing a definition…

37. Mapping the space of Ontology Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Axiomatic Theory GlossaryTaxonomy Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies What is represented? How is it represented?

38. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

39. Mapping the space of Ontology Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies What is represented?

40. Common Principles (1) There is a mind-independent reality of objects in the world (2) There is (human) knowledge or belief Humans use language to exchange information about (1) and (2) Semantically interoperable information systems should be able to deal with language, knowledge and reality

41. Mind Independent Reality Individual Entities (Instances, Particulars) Intelligent Minds triangle square circle trilateral rectangle square circle Natural Language triangle square circle triangle square circle triangle square circle triangle square circle

42. Mind Independent Reality Individual Entities (Instances, Particulars) Intelligent Minds triangle square circle trilateral rectangle square circle Natural Language triangle square circle triangle square circle triangle square circle triangle square circle Classes

43. Examples from Terminological Systems reference to reality reference to knowledge Viral Hepatitis X Viral Hepatitis, diagnosed by clinical examination XX Prevented pregnancy X Diabetes, not elsewhere classified XX Brain transplant X Absent toe X Absence of toe X Relative of possible smoker XX Health Problems related to Mars mission X Yin deficit ?X

44. Mind Independent Reality Individual Entities (Instances, Particulars) RealistConceptualistNominalist Schools of Thinking Categories UniversalsConceptsNames Realist Universals

45. What is represented? How is it represented? Mapping the space of Ontology: Realist perspective Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Universals (types, kinds) are invariants in reality, e.g. cell, molecule, eye, inflammation, … All universals refer to non- empty (at some moment) classes of (individual) entities in the world Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

46. Mind Independent Reality Individual Entities (Instances, Particulars) RealistConceptualistNominalist Schools of Thinking Categories UniversalsConceptsNames Realist Universals Conceptualist Concepts

47. What is represented? How is it represented? Mapping the space of Ontology Conceptualist perspective Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Concepts do not necessarily imply the extension to classes in reality (“retinal transplant”, “yin deficiency”, “missing digit”, “prevented pregnancy”) Concepts as mind constructs may be oriented to prototypes, their extension exhibits large inter-individual variation Concepts can be related by imprecise conceptual relations such as “is broader as” Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

48. RealistConceptualistNominalist Schools of Thinking Mind Independent Reality Categories UniversalsConceptsNames triangle square circle trilateral rectangle square circle Individual Entities (Instances, Particulars) Reference to Reality

49. What is represented? How is it represented? Mapping the space of Ontology Nominalist perspective Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Names are are created in an ad hoc fashion from linguistic predicates. Examples: “People in SR 1048 at 7pm today” “Nontraffic accident involving other off-road motor vehicle ” (ICD9-CM: E821 ) Tuberculosis of lung, bacteriological and histological examination not done (ICD-10: A16.1) “Follow-up inpatient consultation for an established patient which requires at least two of these three key components: a detailed interval history; a detailed examination; medical decision making of high complexity. Counseling and/or coordination of care with other providers or agencies are provided consistent with the nature of the problem(s) and the patient's and/or family's needs. Usually, the patient is unstable or has developed a significant complication or a significant new problem. Physicians typically spend 30 minutes at the bedside and on the patient's hospital floor or unit.” (Current Procedural Terminology Code: HCPT06 ) Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

50. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

51. What is represented? How is it represented? Mapping the space of Ontology Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

52. What is represented? How is it represented? Mapping the space of Ontology Catalogs Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies Hierarchical Ordering of Semantic Nodes

53. What is represented? How is it represented? Mapping the space of Ontology Catalogs Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Catalog: a set of terms without constraints (formal or informal) to characterize their meaning. Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

54. What is represented? How is it represented? Mapping the space of Ontology Glossary Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Glossary: catalogue with glosses in natural language. Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

55. What is represented? How is it represented? Mapping the space of Ontology Taxonomy Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies Taxonomy: terms (and glosses) are organized into a subsumption (subclass) hierarchy with Property inheritance

56. What is represented? How is it represented? Mapping the space of Ontology Thesaurus Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Thesaurus: taxonomy coupled with additional semantic relations (part-of, similar to, etc.). Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

57. What is represented? How is it represented? Mapping the space of Ontology Conceptual Schemas Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Conceptual Schema: Set of terms, attributes and relations (or frames and slots) with explicit descriptions (definitions) Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

58. What is represented? How is it represented? Mapping the space of Ontology Axiomatic Theories Representation restricted to real world entities Representation of term meanings Axiomatic Theory GlossaryTaxonomy Representation of arbitrary propositions Thesaurus Strict Ontologies Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Axiomatic Theory: Formal system with a clear semantics that captures the meaning of the adopted vocabulary via logical formulas. Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

59. Mapping the space of Ontology: Notions of Ontology Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Axiomatic Theory GlossaryTaxonomy Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Perspective of Philosophical Ontology Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

60. Mapping the space of Ontology: Notions of Ontology Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Axiomatic Theory GlossaryTaxonomy Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies AI Perspec- tive

61. Mapping the space of Ontology: Notions of Ontology Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Axiomatic Theory GlossaryTaxonomy Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf Linguistic / Cognitive Perspective Gunnar Klein & Barry Smith: ontology.buffalo.edu/concepts/ ConceptsandOntologies

62. Mapping the space of Ontology: Biomedical Vocabularies Representation restricted to real world entities Representation of term meanings Representation of arbitrary propositions Axiomatic Theory GlossaryTaxonomy Thesaurus Catalog Conceptual Schema from Borgo et al. http://www.loa-cnr.it/Tutorials/ESSLLI1.pdf ICD MeSH SNOMED CT GO WordNet

63. Content A cruise through the O-Space The "O-word": Terminological Clarification Purposes of Ontologies Mapping the O-Space What is represented How is it represented Practice of Good Ontology

64. Don’t mix up universals (Concepts, Classes) with individuals (Instances) subclass-of (Motor Neuron, Neuron ) (FMA, OpenGALEN) is-a (Motor Neuron, Neuron) instance-of (Motor Neuron, Neuron) (FlyBase) But: instance-of (my Hand, Hand) instance-of (this amount of insulin, Insulin) instance-of (Germany, Country) not: instance of (Heart, Organ) not: instance of (Insulin, Protein) is-a = subclass-of: Taxonomic Subsumption instance-of Class Membership

65. Keep in mind the meaning of taxonomic subsumption (is-a) is-a (A, B) = def  x: instance-of(x, A)  instance-of (x, B) is-a (Geographic Area, Spatial Concept) (UMLS-SN) is-a (Spatial Concept, Idea of Concept) (UMLS-SN)  instance-of (Freiburg, Idea or Concept) ?? is-a (Protein Family or Group, Protein, ) (GENIA) ?? No subclassing without inheritance!

66. Don’t use superclasses to express roles is-a (Fish, Animal) is-a (Fish, Food) ?? is-a (Acetylsalicylic Acid, Salicylate) is-a (Acetylsalicylic Acid, Analgetic Drug) ?? Be aware of the “rigidity” of classes

67. Endurant (Continuant) Physical Amount of matter Physical object Feature Non-Physical Mental object Social object … Perdurant (Occurrent) Static State Process Dynamic Achievement Accomplishment Quality Physical Qualities Spatial location … Temporal Q ualitie s Temporal location … Abstract Q ualitie s … Abstract Quality region Time region Space region Color region Source: S. Borgo ISTC-CNR Example: DOLCE’s Upper Ontology Partition the ontology by principled upper level categories

68. Endurants vs. Perdurants Endurants (Continuants): All proper parts are present whenever they are present (wholly presence, no temporal parts) Exist in time Can genuinely change in time May have non-essential parts Need a time-indexed parthood relation Perdurants (Occurrents): Only some proper parts are present whenever the perdurant is present (partial presence,temporal parts ) Happen in time Do not change in time All parts are essential Do not need a time-indexed parthood relation

69. Be aware of ambiguities “Institution” (NCIT) may refer to 1. (abstract) institutional rules 2. (concrete) things instituted 3. act of instituting sth. “Tumor” 1. evolution of a tumor as a disease process 2. having a tumor as a pathological state 3. tumor as a physical object “Gene” 1. a (physical) sequence of nucleotides on a DNA chain 2. a collection of (1) 3. A piece of information conveyed by (1)

70. Use semantically precise Basic Relations Barry Smith, Werner Ceusters, Bert Klagges, Jacob Köhler, Anand Kumar, Jane Lomax, Chris Mungall, Fabian Neuhaus, Alan L Rector and Cornelius Rosse. Relations in biomedical ontologies. Genome Biology, 6(5), 2005.

71. Nontaxonomic Relations between Classes are ambiguous ! has-part(Neuron, Axon) (FMA) Does every neuron has an axon? has-part(Cell, Axon) (Gene Ontology) Do cells without axons exist ? Do axons without cells exist ?

72. Nontaxonomic Relations between Classes are ambiguous ! A, B are classes, inst-of = class membership rel : relation between instances Rel : relation between classes Rel (A, B) = def  x: inst-of (x, A)  inst-of (y, B)  rel (x, y) OR  x: inst-of(x, A)   y: inst-of (y, B)  rel (x, y) OR  x: inst-of(x, A)   y: inst-of (y, B)  rel (x, y) AND  y: inst-of(y, B)   x: inst-of (x, A)  rel (x, y)

73. Nontaxonomic Relations between Classes are ambiguous ! A, B are classes, inst-of = class membership rel : relation between instances Rel : relation between classes Rel (A, B) = def  x: inst-of (x, A)  inst-of (y, B)  rel (x, y) OR  x: inst-of(x, A)   y: inst-of (y, B)  rel (x, y) OR  x: inst-of(x, A)   y: inst-of (y, B)  rel (x, y) AND  y: inst-of(y, B)   x: inst-of (x, A)  rel (x, y)

74. Example: Part-of and Has-Part between Classes

75. MediLOG: Ontology activities Three EU projects SemanticMining: Semantic Interoperability and Data Mining in Biomedicine @neurist: Integrated decision support system to assess the risk of aneurysm rupture in patients and to optimize their treatments. BootSTREP: Integration of biological fact databases and terminological repositories to implement a text analysis system which continuously increases their coverage by analyzing biological documents.

77. MediLOG: Ontology activities Three EU projects SemanticMining: Semantic Interoperability and Data Mining in Biomedicine @neurist: Integrated decision support system to assess the risk of aneurysm rupture in patients and to optimize their treatments. BootSTREP: Integration of biological fact databases and terminological repositories to implement a text analysis system which continuously increases their coverage by analyzing biological documents.

79. MediLOG: Ontology activities Three EU projects SemanticMining: Semantic Interoperability and Data Mining in Biomedicine @neurist: Integrated decision support system to assess the risk of aneurysm rupture in patients and to optimize their treatments. BootSTREP: Integration of biological fact databases and terminological repositories to implement a text analysis system which continuously increases their coverage by analyzing biological documents.

81. Thank you for your attention

82. Conclusions Many ontology builders and users have a implicitly promiscuous attitude both towards philosophical categories and formal properties Different application contexts require different styles of ontologies / terminological systems. There is not the one purist approach Ontology builders should avoid pitfalls by adhering to best practice guidelines Every terminology system should have its defined place in the “ontological space” and have a clear commitment of what it represents and how This is an important requirement for reference terminologies such as SNOMED