Medical%20Ontologies:%20An%20Overview

1
Medical Ontologies An Overview

• Barry Smith
• http//ifomis.de
• January 2004

2
IFOMIS

• Institute for Formal Ontology and
• Medical Information Science
• Faculty of Medicine
• University of Leipzig

3
Partners

• Laboratory for Applied Ontology, Trento and Rome
• Language  Computing nv, Zonnegem, Belgium
• Ontology Works, Baltimore
• Structural Informatics Group, Department of
  Biological Structure, University of Washington,
  Seattle, USA
• Cognitive Science Laboratory, Princeton University

4
Three levels of ontology

• formal (top-level) ontology dealing with
  categories employed in every domain
• object, event, whole, part, instance, class
• 2) domain ontology, applies top-level system to
  a particular domain
• cell, gene, drug, disease, therapy
• 3) terminology-based ontology
• large, lower-level system
• Dupuytrens disease of palm, nodules with no
  contracture

5
Three levels of ontology

• formal (top-level) ontology dealing with
  categories employed in every domain
• object, event, whole, part, instance, class
• 2) domain ontology, applies top-level system to
  a particular domain
• cell, gene, drug, disease, therapy
• 3) terminology-based ontology
• large, lower-level system
• Dupuytrens disease of palm, nodules with no
  contracture

6
Three levels of ontology

• formal (top-level) ontology dealing with
  categories employed in every domain
• object, event, whole, part, instance, class
• 2) domain ontology, applies top-level system to
  a particular domain
• cell, gene, drug, disease, therapy
• 3) terminology-based ontology
• large, lower-level system
• Dupuytrens disease of palm, nodules with no
  contracture

7
IFOMIS

• Institute for Formal Ontology and Medical
  Information Science
• Leipzig
• http//ifomis.de
• philosophers and medical informaticians
  attempting to build and test a Basic Formal
  Ontology for applications in biomedical and
  related domains

8
IFOMIS

• use basic principles of philosophical ontology
• for quality assurance and alignment of
  biomedical ontologies

9
Compare

1. pure mathematics (theories of structures such as
   order, set, function, mapping) employed in every
   domain
2. applied mathematics, applications of these
   theories re-using the same definitions,
   theorems, proofs in new application domains
3. physical chemistry, biophysics, etc. adding
   detail

10
Three levels of ontology
?????

• formal (top-level) ontology
• medical ontology has nothing like the technology
  of definitions, theorems and proofs provided by
  pure mathematics
• 2) domain ontology
• UMLS Semantic Network, GALEN CORE
• 3) terminology-based ontology
• UMLS, SNOMED-CT, GALEN, FMA

11
Strategy

• Part 1 Provide an overview of medical ontologies
  and of the top-level ontologies which they
  implicitly define
• Part 2 Show how principles of classification and
  definition derived from top-level ontology can
  help in quality assurance of terminology-based
  ontologies and in ontology alignment
• Part 3 The Gene Ontology
• Part 4 Medical Fact Net

12
(No Transcript)
13
UMLS Semantic Network

• entity event
• physical conceptual
• object entity

14
UMLS Semantic Network

• entity event
• physical conceptual
• object entity

15
conceptual entity

• Organism Attribute
• Finding
• Idea or Concept
• Occupation or Discipline
• Organization
• Group
• Group Attribute
• Intellectual Product
• Language

16

• 
  conceptual
• entity
• idea or concept
• functional concept
• body system

17

• entity
• physical conceptual
• object entity
• idea or concept
• functional concept
• body system

confusion of entity and concept
18
Functional Concept

• Body system is_a Functional Concept.
• but
• Concepts do not perform functions or have
  physical parts.

19
This
is not a concept
20
The Hydraulic Equation

• BP COPVR
• arterial blood pressure is directly proportional
  to the product of blood flow (cardiac output, CO)
  and peripheral vascular resistance (PVR)

21
Confusion of Ontology and Epistemology

• blood pressure is an Organism Function,
• cardiac output is a Laboratory or Test Result or
  Diagnostic Procedure
• BP COPVR thus asserts that
• blood pressure is proportional either to a
  laboratory or test result or to a diagnostic
  procedure

22
entities

• independent dependent occurrents
• continuants continuants (always
  dependent)
• ORGANISMS ROLES PROCESSES
• CELLS FUNCTIONS HISTORIES
• MOLECULES CONDITIONS LIVES
  (diseases) (courses of
• 
  diseases)

23
entities

• independent dependent occurrents
• continuants continuants (always
  dependent)
• ORGANISMS ROLES PROCESSES
• CELLS FUNCTIONS HISTORIES
• MOLECULES CONDITIONS LIVES
  (diseases) (courses of
• 
  diseases)

classes
instances
24
A three-category ontology along these lines
accepted by

• DOLCE first module of Semantic Web Wonderweb
  Foundational Ontologies Library
• BFO IFOMIS Basic Formal Ontology
• LC LinKBase
• UMLS-SN
• Gene Ontology

25
(No Transcript)
26
Principles for Building Medical Ontologies
27
Examples

• Dont confuse entities with concepts
• Dont confuse domain entities with logical or
  computational structures
• Dont confuse ontology with epistemology
• Dont confuse is_a with has_role

28
Further Principles

• univocity terms should have the same meanings
  (and thus point to the same referents) on every
  occasion of use
• UMLS-SN
• organization body plan
• organization social organization

29
univocity

• Gene Ontology
• part_of can be part of (flagellum part_of
  cell)
• part_of is sometimes part of (replication
  fork part_of the nucleoplasm)
• part_of is included as a sublist in

30
dont forget instances

• part_of as a relation between classes
• vs. part as a relation between instances
• A part_of B
• every instance of A is part of some instance of B
  
• every instance of B has some instance of A as part

31
Part_of as a relation between classes is more
problematic than is standardly supposed

• testis part_of human being ?
• heart part_of human being ?

32
objectivity

• which classes exist is not a function of our
  biological knowledge.
• (Terms such as unknown or unclassified or
  unlocalized do not designate biological natural
  kinds.)
• GO
• aminoadipate-semialdehyde dehydrogenase complex
  is_a unlocalized

33
rules for definitions

• intelligibility the terms used in a definition
  should be simpler (more intelligible) than the
  term to be defined
• definitions do not confuse definitions with the
  communication of new knowledge

34
substitutability

• in all so-called extensional contexts a defined
  term should be substitutable by its definition in
  such a way that the result is both grammatically
  correct and has the same truth-value as the
  sentence with which we begin
• GO0015070 toxin activity
• Definition Acts as to cause injury to other
  living organisms.

35
substitutability

• There is toxin activity here
• There is acts as to cause injury to other living
  organisms here

36
(No Transcript)
37
GO the Gene Ontology

• 3 large telephone directories of standardized
  designations for gene functions and products
• organized into hierarchies via is_a and part_of

38
GO

• can in practice be used only by trained
  biologists (with know how)
• whether a GO-term truly stands in the is_a
  relation depends e.g. on the type of organism
  involved
• glycosome is part-of cytoplasm only for
  Kinetoplastidae
• Computers have no counterpart of such
  context-dependent know-how

39
GO divided into three disjoint term hierarchies

• the cellular component ontology,
• e.g. flagellum, chromosome, cell
• the molecular function ontology,
• e.g. ice nucleation, binding, protein
  stabilization
• the biological process ontology,
• e.g. glycolysis, death

40
Primary aim of GO

• not rigorous definition and principled
  classification
• but rather providing a practically useful
  framework for keeping track of the biological
  annotations that are applied to gene products

41
Thesis 1

• With increasing size, GO will be required to
  increase the degree to which it is a controlled
  vocabulary which satisfies not merely the needs
  of human biologists but also the needs of
  automatic consistency-checking and updating
  systems

42
Thesis 2

• GO can realize its goal more adequately (and
  avoid many coding errors) by taking ontology
  (especially the logic of classifications and
  definitions) seriously

43
GO the Gene Ontology

• GO divided into 3 separate hierarchies each
  organized via is_a and part_of

44
Problems with is_a

• A is_a B every instance of A is an instance of B

45
Problems with is_a

• Holliday junction helicase complex is_a
• unlocalized
• protein storage vacuole is_a
• vacuole (sensu Streptophyta)
• R7 differentiation is_a eye photoreceptor
  differentiation (sensu Drosophilia).

46
Uses of part_of

• membrane part-of cell, intended to mean a
  membrane is a part-of any cell
• flagellum part-of cell, intended to mean a
  flagellum is part-of some cells
• replication fork part-of cell cycle, intended
  to mean a replication fork is part-of the
  nucleoplasm only during certain times of the cell
  cycle
• regulation of sleep part-of sleep, should be
  corrected to regulation of sleep is co-located
  with and is causally involved with the sleep
  process.

47
Problems with part_of

• part_of can be part of (flagellum part_of
  cell)
• part_of is sometimes part of (replication
  fork part_of the nucleoplasm)
• part_of is included as a sublist in

48
Problems with GO Molecular Functions

• anti-coagulant activity (defined as a
  substance that retards or prevents coagulation)
• enzyme activity (defined as a substance that
  catalyzes)
• structural molecule (defined as the action of
  a molecule that contributes to structural
  integrity)

49
GO0005199 structural constituent of cell wall

• Definition The action of a molecule that
  contributes to the structural integrity of a cell
  wall.
• confuses actions, which GO includes in its
  function ontology, with constituents, which GO
  includes in its cellular component ontology

50

• extracellular matrix structural constituent
• puparial glue (sensu Diptera)
• structural constituent of bone
• structural constituent of chorion (sensu Insecta)
  
• structural constituent of chromatin
• structural constituent of cuticle
• structural constituent of cytoskeleton
• structural constituent of epidermis
• structural constituent of eye lens
• structural constituent of muscle
• structural constituent of myelin sheath
• structural constituent of nuclear pore
• structural constituent of peritrophic membrane
  (sensu Insecta)
• structural constituent of ribosome
• structural constituent of tooth enamel
• structural constituent of vitelline membrane
  (sensu Insecta)

51
Why do these problems arise?

• Because GO has no clear formal understanding of
  the role of temporal relations in organizing an
  ontology
• (thus also no clear understanding of the
  difference between a function and the activity
  which is the realization of a function GO runs
  these two together)

52
As GO increases in size and scope

• it will be increasingly difficult to maintain
  the semantic consistency we desire without
  software tools that perform consistency checks
  and controlled updates.
• The addition of each new term will require the
  curator to understand the entire structure of GO
  in order to avoid redundancy and to ensure that
  all appropriate linkages are made with other
  terms.

53
Problems with GOs compositionality
sensu / with from in resulting regulating regulation of complex constituting constitution
54
/

• GO0008608 microtubule/kinetochore interaction
• df Physical interaction between microtubules and
  chromatin via proteins making up the kinetochore
  complex,
• GO0001539 ciliary/flagellar motility
• df Locomotion due to movement of cilia or
  flagella.

55
/

• GO0045798 negative regulation of chromatin
  assembly/disassembly
• df Any process that stops, prevents or reduces
  the rate of chromatin assembly and/or disassembly
• GO0000082 G1/S transition of mitotic cell cycle
• defined as Progression from G1 phase to S phase
  of the standard mitotic cell cycle.

56
/

• GO0001559 interpretation of nuclear/cytoplasmic
  to regulate cell growth
• df The process where the size of the nucleus
  with respect to its cytoplasm signals the cell to
  grow or stop growing.

57
/

• GO0015539 hexuronate (glucuronate/galacturonate)
  porter activity
• df Catalysis of the reaction hexuronate(out)
  cation(out) hexuronate(in) cation(in)

58
Problems with GOs consistency

• GO 0030430 host cell cytoplasm part-of GO018995
  host
• host cell cytoplasm df The cytoplasm of a host
  cell.
• host df Any organism in which another organism,
  especially a parasite or symbiont, spends part or
  all of its life cycle and from which it obtains
  nourishment and/or protection.

59
Cellular Component

• Another problem with host
• It is not a cellular component (and not a
  molecular function, and not a biological process,
  either)
• GO has adult walking behavior
• but not adult or walking
• GO has eye pigmentation but not eye

60
Solution

• Link GO to external ontologies
• of organism types (to solve the sensu problem)
• of anatomy, to solve the eye problem
• of coarse medical reality, to solve the adult
  walking behavior problem) (see MFN below)

61
note that such linkages are possible

• only if GO itself has a coherent formal
  architecture

62
(No Transcript)
63
Medical Fact Net

• Medical Belief Net (MBN)
• large, heterogeneous, open-source corpus of
  medical sentences in the English language
  expressed in the form of grammatically complete
  statements and assessed by the degree to which
  they are understandable and assented to by
  typical non-expert human subjects.
• Medical Fact Net (MFN) subclass of MBN
  receiving high marks on the scale of correctnesss
  from medical experts
• MFN intersection of non-expert beliefs about
  medical phenomena and truths validated by medical
  experts.

64
Medical Word Net

• lexical database extending the Princeton
  WordNet by all the medical terms encountered in
  MBN
• First in (US) English
• Then in German
• First for adults, then for children
• First for medicine, then for

65
MBN/MFN/MWN Formal Architecture

• Semi-automatically generated graph-based parsing
  of each sentence
• formal ontology of all MFN entities and
  relationships
• mapping into the UMLS Metathesaurus.

66
Evaluation

• MFN will be integrated into an existing
  term-search-based on-line consumer health portal
  based in such a way that MFN sentences are used
  to direct users to information sources. We will
  then measure the degree to which this results in
  greater user satisfaction by setting up an
  experiment in which customers of the portal are
  randomly assigned to one of two groups one to
  which access to MFN is offered, and other for
  which simple term-searching is used.

67
Significance

• Non-expert language of family members, advisors,
  administrators, nurses, paramedics, lawyers
• Research on differences between everyday language
  and technical language

68
Mismatches in Doctor-Patient Communication

• Question Text My seven-year-old son developed a
  rash today that I believe to be chickenpox. My
  concern is that a friend of mine had her
  10-day-old baby at my home last evening before we
  were aware of the illness. Is there cause for
  concern at this point?
• Answer Text Chickenpox is the common name for
  varicella infection. ... You are correct in
  that a person with chickenpox can be contagious
  for 48 hours before the first vesicle is seen.
  ...

69
Non-Expert Language in Online Communication

• Need to integrate free text and structured data.
• E-health services need automatic ways to respond
  to questions in standard forms, and to provide
  internet-accessible medical knowledge that is
  both reliable and accessible to the non-expert.

70
Diagnostic decision support

• we might associate collections of utterances
  stored in MBN describing symptoms sourced to
  single patients with metadata recording
  subsequent diagnosis. Trained on this corpus, the
  system could establish patterns of association
  between specific sequences of utterances and
  specific diseases one could then test the degree
  to which such associations are sufficiently
  strong as to produce usable automatic diagnosis
  on the basis of patient inputs.

71
Medical education/medical literacy

• Use MBN to evaluate of the reliability of the
  medical knowledge of different non-expert
  communities.
• Use MFN to develop tools to support face-to-face
  education of lay people in the fields of medicine
  and health care
• MBN provides opportunities for a new type of
  research in the field of consumer health.
• e.g. on basic kinds in the medical domain à la
  Eleanor Rosch

72
Medical Coverage in WordNet 2.0

• WordNets coverage of domains like medicine,
  physics, and geology is very limited.
• coverage of medical terms represents a mixture
  of folk and expert vocabulary.

73
MFN From Words to Facts

• Do for (non-expert) medicine what Belsteins Fact
  Database does for (expert) Biochemistry
• Relation to CYC
• Relation to FrameNet
• Botany Knowledge Base
• DARPAs Rapid Knowledge Formation project.

74
Sources

• Lexical knowledge bases, such as
• the relevant general lexical information
  contained in WordNet
• lexical knowledge-bases of lay medical vocabulary
  
• medical dictionaries and large medical
  terminology and ontology systems such as the UMLS
  Specialist Lexicon, the Foundational Model of
  Anatomy
• Statement or fact knowledge bases, such as
• d. open-source linguistic corpora, public health
  documents, internet resources
• e. the relevant example sentences in the FrameNet
  and WordNet corpora
• f. free text sources
• g. the results of transforming the content of
  lexical knowledge bases (especially WordNet) into
  statements

75
Generation from lexical databases

• treat a database like WordNet or LinKBase as a
  set of links tLt', between terms (where L ranges
  over 'is-a', 'part-of', 'is-caused-by', etc.).
• We form the subset of this set by restricting
  the values of t and t' to those which terms occur
  in MWN
• Some members of the resulting class of tLt'
  formula can then be transformed into English
  sentences automatically. For example each t is-a
  t'-formula can be transformed into a sentence of
  the form ' a t is a type of t' '
• Other tLt' formula can be converted by hand into
  English sentences, for example "forearm
  HAS-PARTIAL-MATERIAL-OVERLAP wrist" can be
  transformed into "the forearm overlaps with the
  wrist" and "the wrist overlaps with the forearm".

76
Problems to be Addressed

• generic medical knowledge of (non-expert) adults

77
Genericity

• Much generic medical knowledge relates to what
  holds for the most part or in most cases or in a
  statistically significant fraction of cases
  (consider smoking causes cancer).

78
Medical knowledge

• is intertwined with knowledge of other domains
• (things that can be involved in an accident )

79
Knowledge

• Much medical knowledge of experts and
  non-experts alike takes the form of knowledge of
  specific cases (Aunt Marys arthritis is always
  worse in the winter).
• MFN should be a repository of medical knowledge
  that is generic and context-independent, the
  counterpart of the theoretical knowledge of the
  sciences.
• Note that lexical knowledge of the sort stored
  in WordNet, too, is both generic and
  context-independent.

80
Expertise

• a crisp separation of expert and non-expert
  sentences is impossible.
• Viagra, anthrax, HIV, Prozac, SARS
• ? experimental design needed to avoid artifacts

81
Completeness

• Problem elementary facts People have two eyes.
  Babies are born. Arms move.
• WordNet contains some coverage particularly of
  elementary facts of the A is type/part of B form
  in virtue of their specific formal architectures
• WordNet synsets can be used to generate long
  lists of elementary facts from single starting
  points

82
Six

• Transform MWN into a large corpus of generic
  beliefs by turning WordNet on its side that is
  we transform a relation such as t1, , tn IS-A
  t1, , tm into n x m sentences of the form
  ti IS-A tk
• and impose filters

83
A New Kind of Linguistics

• MFN part and parcel of recent attempts in the
  biomedical sciences to confront problems of
  similar scope in the development of large
  fact-repositories such as KEGG or Swiss-Prot.
• In its final form it should be consistent with
  the knowledge that is contained also in other
  fact repositories both at the expert and the
  non-expert level and serve to integrate them
  together in a federated database.

84
Adult walking behavior

• will be freed from its lonely status inside GO

85

• The End