600 likes | 622 Views
Explore the importance of relations in Semantic Network for biomedical ontologies, emphasizing the significance of inheritance and key relations like 'part_of' and 'contains'. Learn about challenges in establishing uniform, significant relations for automated inference and their impact on biomedical ontology integration. Discover ways to bridge the "Granularity Gulf" in reasoning and enforce coherent definitions to facilitate automatic reasoning across different data granularities in healthcare settings.
E N D
Semantic Network (SN) and Biomedical Ontology Barry Smith Department of Philosophy, University at Buffalo Institute for Formal Ontology and Medical Information Science ifomis.org
Assumption • SN is designed to support automatic reasoning involving multiple UMLS source terminologies • Conclusion: • Relations in SN are very important
Inheritance • Body part, Organ or Organ Component location_of Biologic Function • Therefore • Body Part, Organ or Organ Component location_of Disease or Syndrome • Alexa: “We can sometimes infer ... you have to bring some medical knowledge to bear”
Part_of as a relation between classes is more problematic than is standardly supposed • testis part_of human being ? • heart part_of human being ?
What we need for automatic inference is uniformly (necessarily) significant relations • A is_a B • Every instance of A is an instance of B
What we need for automatic inference is uniformly (necessarily) significant relations • A is_a B • Every instance of A is an instance of B • We hope all is_a relations are exceptionless in this sense
Somenon-is_a relations are exceptionless in this sense • Fully formed anatomical structure contains body substance
But most are not • Bacterium causes Experimental Model of Disease • Experimental Model of Disease affects Fungus • Experimental model of diseaseis_a Pathologic Function
Bacterium causes Experimental Model of Disease • causes– Brings about a condition or an effect. Implied here is that an agent, such as for example, a pharmacologic substance or an organism, has brought about the effect. This includes induces, effects, evokes, and etiology.
GALEN:Vomitus contains carrot • Gene Ontology:Menopause part_of Death • HL7: Individual Allele is_a Act of Observation
Thesis: Biomedical ontology integration • will never be achieved through integration of meanings or concepts in people’s heads • the problem is precisely that different user communities use different concepts
Promise of evidence-based medicine in the genomics era:integrating biomedical terminologies with EHR data • need facility for dealing with time and instances (particulars, actual cases) with this tumor here and now in this breast ...
Move from associative relations between meanings to strictly defined relations between the entities themselves • See: • Smith, Ceusters, Klagges, Köhler, Kumar, Lomax, Mungall, Neuhaus, Rector, Rosse • “Relations in Biomedical Ontologies” • Genome Biology, in press
Clear instructions • Fewer mistakes
Key idea • To define ontological relations like SN’s • part_of, contains, adjacent-to • we need also to take account of instances and time • (= link to Electronic Health Record)
Kinds of relations • <class, class>: is_a, part_of, ... • <instance, class>: this explosion instance_of the class explosion • <instance, instance>: Mary’s heart part_of Mary
Kinds of relations • <class, class>: is_a, part_of, ... • <instance, class>: this explosion instance_of the class explosion • <instance, instance>: Mary’s heart part_of Mary • = instance-level part_ofis a primitive (you can’t define everything, on pain of circularity)
part_of • A part_of B =def. • for all a and all t, • if a is an instanceof A at time t, • then there is some instance b of B • such that a is an instance-level part_ofb at t • ALL-SOME STRUCTURE
part_of • A part_of B =def. • for alla and allt, • if a is an instanceof A at time t, • then there is someinstance b of B • such that a is an instance-level part_ofb at t • ALL-SOME STRUCTURE
testis part_of human being - NO • human testis part_of human being - YES • human ovary part_of human being - YES
same instance C1 C c att c att1 time transformation_of mature RNAtransformation_of pre-RNA fetus transformation_ofembryo adulttransformation_of child
transformation_of • C2 transformation_of C1 =def. any instance of C2 was at some earlier time an instance of C1
Note the problem of inverses here • Not every child becomes transformed into an adult
The Granularity Gulf as an obstacle to reasoning • most existing data-sources are of fixed, single granularity • many (all?) clinical phenomena cross granularities
C1 C c att c att1 embryological development
tumor development C1 C c att c att1
Advantages of the methodology of enforcing commonly accepted coherent definitions • promote quality assurance (better coding) • guarantee automatic reasoning across ontologies and across data at different granularities, from molecule to clinic • yields direct connection to times and instances in EHR
Automatic reasoning • non-is_a relations are all-some relations • A R B =def for all instances a of A there is some instance b of B such that a r b • where r is some instance-level relation • If you know A R B, and you know that a is an instance of A, then you knowthat there is some instance b of B • and inheritance is unrestrained (exceptionless) • if you know B R C you can reason with this instance b to infer that there is some C, • and so on
Conclusions for SN • Remove the merely ‘possibly significant relations’ (these are less than facts) • Reform definitions (remove circularity) • Remove those relations, such as prevents which cannot be given a coherent instance-based all-some definition • Reform treatment of inverses
prevents • Definition: Stops, hinders or eliminates an action or condition.Inverse: prevented_by • contraception prevents pregnancy • pregnancy prevented_by contraception
Better treatment of prevention • contraception • causes • prevention_of_pregnancy
Reform treatment of inverses • adjacent_to – “Close to, near or abutting another physical unit with no other structure of the same kind intervening. This includes adjoins, abuts, is contiguous to, is juxtaposed, and is close to.” • Inverse: adjacent_to
Adjacent_to is not its own inverse • nuclear membrane adjacent_to cytoplasm • BUT NOT: • cytoplasm adjacent_to nuclear membrane • ovary adjacent_to parietal pelvic peritoneum • BUT NOT: • parietal pelvic peritoneum adjacent_to ovary
Better treatment of inverses • Use preceded_by • not precedesas primary relation • preceded_bysupports inheritance • (supports automatic reasoning) • embryological development precedes birth NOT EXCEPTIONLESS
If NLM does not reform SN • in something like this way, then someone else will build a competitor to integrate the UMLS for purposes of automatic reasoning and integration across granularities
http://ifomis.org • The End
Human-Caused Phenomenon or Process (Environmental Effect of Humans): • Phenomenon and Process put together
UMLS Semantic Types Entity Event Physical Object Conceptual Entity Phenomenon or Process Activity
genepart_ofcell component • body systemconceptual_part_of • fully formed anatomical structure
conceptual • entity • idea or concept • functional concept • body system
But: • Gene or Genome is defined as: “A specific sequence … of nucleotides along a molecule of DNA or RNA …” • and • nucleotide sequence is_a conceptual entity
entity • physical conceptual • object entity • idea or concept • functional concept • body system confusion of entity and concept
Functional Concept: • Body system is_a Functional Concept. • but: • Concepts do not perform functions or have physical parts.
This: is not a concept
UMLS-SN Semantic Relation • producesDefinition: Brings forth, generates or creates. Inverse: produced_by • artificial insemination produces pregnancy • pregnancy produced by artificial insemination
Definitions • conceptual_part_of – Conceptually a portion, division, or component of some larger whole. • should not be circular
part_of – “Composes, with one or more other physical units, some larger whole. This includes component of, division of, portion of, fragment of, section of, and layer of.” Inverse: has-part • contains – “Holds or is the receptacle for fluids or other substances. This includes is filled with, holds, and is occupied by.” Inverse: contained_in • consists_of – “Is structurally made up of in whole or in part of some material or matter. This includes composed of, made of, and formed of.” Inverse: constitutes • connected_to – “Directly attached to another physical unit as tendons are connected to muscles. This includes attached to and anchored to.” Inverse: connected_to • interconnects – “Serves to link or join together two or more other physical units. This includes joins, links, conjoins, articulates, separates, and bridges.” Inverse: interconnected by • branch_of – “Arises from the division of. For example, the arborization of arteries.” Inverse: has_branch
tributary_of – “Merges with. For example, the confluence of veins.” Inverse: has_tributary • ingrediant_of – “Is a component of, as in a constituent of a preparation.” Inverse: has_ingredient • physically_related_to – “Related by virtue of some physical attribute or characteristic.” Inverse: physically_related_to