Aligning relationships in the UMLS Methods and preliminary

Aligning relationships in the UMLSAligning relationships in the UMLSMethods and preliminary resultsMethods and preliminary results

Olivier BodenreiderOlivier Bodenreider

Lister Hill National CenterLister Hill National Centerfor Biomedical Communicationsfor Biomedical CommunicationsBethesda, Maryland Bethesda, Maryland -- USAUSA

New JerseyNew Jersey’’s Sciences Scienceand Technology Universityand Technology UniversityComputer Science DepartmentComputer Science DepartmentOctober 23, 2007October 23, 2007

2Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical Communications

AcknowledgmentsAcknowledgments

AlexaAlexa T. McCrayT. McCrayHarvard Medical SchoolHarvard Medical SchoolBoston, MassachusettsBoston, Massachusetts

Lowell Lowell VizenorVizenorOntologyWorksOntologyWorks, Inc, IncBaltimore, MarylandBaltimore, Maryland


OverviewOverview

The UMLS: A twoThe UMLS: A two--level structurelevel structureAligning relationshipsAligning relationships

Multiple approachesMultiple approachesPreliminary resultsPreliminary results

Towards an ontology of relationshipsTowards an ontology of relationships

The UMLSThe UMLS

A twoA two--level structurelevel structure


Unified Medical Language SystemUnified Medical Language System

SPECIALIST LexiconSPECIALIST Lexicon360,000 lexical items360,000 lexical itemsPart of speech and variant informationPart of speech and variant information

MetathesaurusMetathesaurus6M names from over 100 terminologies6M names from over 100 terminologies1.5M concepts1.5M concepts8M relations8M relations

Semantic NetworkSemantic Network135 high135 high--level categorieslevel categories7000 relations among them7000 relations among them

Lexicalresources

Ontologicalresources

Terminologicalresources

Heart

Concepts

Metathesaurus

38

237

49

5

16

13 22

Esophagus

Left PhrenicNerve

HeartValves

FetalHeart

Medias-tinum

SaccularViscus

AnginaPectoris

CardiotonicAgents

TissueDonors

AnatomicalStructure

Fully FormedAnatomical

StructureEmbryonicStructure

Body Part, Organ orOrgan Component Pharmacologic

Substance

Disease orSyndrome

PopulationGroup

Semantic Types

SemanticNetwork

UMLS Metathesaurus


Source VocabulariesSource Vocabularies

143 source vocabularies143 source vocabularies17 languages17 languages

Broad coverage of biomedicineBroad coverage of biomedicine5.9M names5.9M names1.4M concepts1.4M concepts8M relations8M relations

Common presentationCommon presentation

(2007AB)


Metathesaurus Metathesaurus Basic organizationBasic organization

ConceptsConceptsSynonymous terms are clustered into a conceptSynonymous terms are clustered into a conceptProperties are attached to concepts, e.g.,Properties are attached to concepts, e.g.,

Unique identifierUnique identifierDefinitionDefinition

RelationsRelationsConcepts are related to other conceptsConcepts are related to other conceptsProperties are attached to relations, e.g.,Properties are attached to relations, e.g.,

Type of relationshipType of relationshipSourceSource


Integrating subdomainsIntegrating subdomains

Biomedicalliterature


MeSH

Genomeannotations

Genomeannotations

GOModelorganisms

Modelorganisms

NCBITaxonomy

Geneticknowledge bases


OMIM

Clinicalrepositories


SNOMED CTOthersubdomains

Othersubdomains

…

AnatomyAnatomy

FMA

UMLS


Integrating subdomainsIntegrating subdomains



Genomeannotations

Genomeannotations

Modelorganisms

Modelorganisms





Othersubdomains

Othersubdomains

AnatomyAnatomy


TransTrans--namespace integrationnamespace integration

Genomeannotations

Genomeannotations

GOModelorganisms

Modelorganisms

NCBITaxonomy



OMIMOther

subdomainsOther

subdomains

…

AnatomyAnatomy

FMA

UMLSAddison Disease (D000224)

Addison's disease (363732003)



MeSH



SNOMED CT

UMLSC0001403


AddisonAddison’’s Disease: s Disease: ConceptConcept

Addison’s Disease

C0001403

ADRENAL INSUFFICIENCY (ADDISON'S DISEASE) ADRENOCORTICAL INSUFFICIENCY, PRIMARY FAILURE Hypoadrenalisms, PrimaryMelasma addisoniiPrimary adrenal deficiency Asthenia pigmentosaBronzed disease Insufficiency, adrenal primary Primary adrenocortical insufficiency Addison's, disease

Maladie d'Addison - FrenchAddison-Krankheit - GermanMorbo di Addison - ItalianDoença de Addison - PortugueseАДДИСОНОВА БОЛЕЗНЬ - Russianアジソン病 - Japanese

An adrenal disease characterized by the progressive destruction of the adrenal cortex, resulting in insufficient production of aldosterone and hydrocortisone. Clinical symptoms include anorexia; nausea; weight loss; muscle ewakness; and hyperpigmentation of the skin due to increase in circulating levels of ACTH precursor hormone which stimulates melanocytes.

Disease or Syndrome

SNOMED CTSNOMED IntlMeSHMedDRA…


Metathesaurus Metathesaurus Evolution over timeEvolution over time

Concepts never die (in principle)Concepts never die (in principle)CUIs are permanent identifiersCUIs are permanent identifiers

What happens when they do die (in reality)?What happens when they do die (in reality)?Concepts can merge or splitConcepts can merge or splitResulting in new concepts and deletionsResulting in new concepts and deletions

Addison's diseaseC0001403

Addison's disease, NOS C0271735

1992 1993 1994 1995 1996 1997 1998 1999 2007…


Metathesaurus Metathesaurus RelationshipsRelationships

Symbolic relations:Symbolic relations: ~8 M pairs of concepts~8 M pairs of conceptsStatistical relations :Statistical relations : ~6 M pairs of concepts ~6 M pairs of concepts (co(co--occurring concepts)occurring concepts)Mapping relations:Mapping relations: ~150,000~150,000

Categorization: Relationships between concepts Categorization: Relationships between concepts and semantic types from the Semantic Networkand semantic types from the Semantic Network


Symbolic relationsSymbolic relations

RelationRelationPair of Pair of ““atomatom”” identifiersidentifiersTypeTypeAttribute (if any)Attribute (if any)List of sources (for type and attribute)List of sources (for type and attribute)

Semantics of the relationship:Semantics of the relationship:defined by its defined by its typetype [and [and attributeattribute]]Recorded Recorded bidirectionallybidirectionally


Symbolic relationships Symbolic relationships TypeType

HierarchicalHierarchicalParent / ChildParent / ChildBroader / Narrower thanBroader / Narrower than

Derived from hierarchiesDerived from hierarchiesSiblings (children of parents)Siblings (children of parents)

AssociativeAssociativeOtherOther

Various flavors of nearVarious flavors of near--synonymysynonymySimilarSimilarSource asserted synonymySource asserted synonymyPossible synonymyPossible synonymy

PAR/CHDRB/RN

SIB

RO

RLSYRQ


Symbolic relationships Symbolic relationships AttributeAttribute

HierarchicalHierarchicalisaisa (is(is--aa--kindkind--of)of)partpart--ofof

AssociativeAssociativelocationlocation--ofofcausedcaused--bybytreatstreats……

CrossCross--references (mapping)references (mapping)


Categorize conceptsCategorize concepts

HighHigh--level categories level categories (semantic types)(semantic types)Assigned by the Assigned by the Metathesaurus editorsMetathesaurus editorsIndependently of the Independently of the hierarchies in which these hierarchies in which these concepts are locatedconcepts are located

Disease or Syndrome

Endocrine Diseases

Adrenal Gland Diseases

Addison’s Disease

Diseases

Adrenal Gland Hypofunction


Categorization vs. hierarchiesCategorization vs. hierarchies

Semantic typeSemantic typeList of all conceptsList of all conceptshaving this semantic typehaving this semantic type

ConceptConceptList of all descendantsList of all descendants

UMLS Semantic Network


Semantic NetworkSemantic Network

Semantic types (135)Semantic types (135)tree structuretree structure2 major hierarchies2 major hierarchies

EntityEntity–– Physical ObjectPhysical Object–– Conceptual EntityConceptual Entity

EventEvent–– ActivityActivity–– Phenomenon or ProcessPhenomenon or Process


Semantic NetworkSemantic Network

Semantic network relationships (54)Semantic network relationships (54)hierarchical (isa = is a kind of)hierarchical (isa = is a kind of)

among typesamong types–– AnimalAnimal isaisa OrganismOrganism–– EnzymeEnzyme isaisa Biologically Active SubstanceBiologically Active Substance

among relationsamong relations–– treats treats isaisa affectsaffects

nonnon--hierarchicalhierarchicalSign or SymptomSign or Symptom diagnosesdiagnoses Pathologic FunctionPathologic FunctionPharmacologic SubstancePharmacologic Substance treatstreats Pathologic FunctionPathologic Function


““Biologic FunctionBiologic Function”” hierarchy (isa)hierarchy (isa)

Biologic Function

Pathologic FunctionPhysiologic Function

Disease orSyndrome

Cell orMolecular

Dysfunction

ExperimentalModel ofDisease

OrganismFunction

Organor TissueFunction

CellFunction

MolecularFunction

Mental orBehavioral

Dysfunction

NeoplasticProcess

MentalProcess

GeneticFunction


Associative (nonAssociative (non--isa) relationsisa) relationsOrganism

process of

EmbryonicStructure

AnatomicalAbnormality

CongenitalAbnormality

AcquiredAbnormality

Fully FormedAnatomicalStructure

AnatomicalStructure

part of

OrganismAttribute

property of

BodySubstance

contains,produces

conceptualpart of

evaluation of

Body Systemconceptualpart of

part of

Body Part, Organ orOrgan Component

part of

Tissue

part of

Cell

part of

CellComponent

Gene orGenome

Body Spaceor Junction

adjacent to

location of

location of

evaluation ofFinding

Laboratory orTest Result

Sign orSymptom

BiologicFunction

PhysiologicFunction

PathologicFunction

Body Locationor Region

conceptualpart of

conceptualpart of

Injury orPoisoning

disrupts

disrupts

co-occurs with


Relationship hierarchyRelationship hierarchy

associated with

functionallyrelated to

temporallyrelated to

physicallyrelated to

spatiallyrelated to

conceptuallyrelated to

associated with


part of

consists of

contains

connected to

interconnects

branch of

tributary of

ingredient of

spatiallyrelated to

location of

adjacent to

surrounds

traverses


co-occurs with

precedes



associated with


spatiallyrelated to




affects

brings about

performs

occurs in

uses

manifestation of

indicates

result of

manages

treats

disrupts

complicates

interacts with

prevents

produces

causes

carries out

exhibits

practices

process of

associated with


spatiallyrelated to




degree of

analyzes

measurement of

measures

diagnoses

property of

derivative of

method of

conceptual part of

issue in

assesses effect of

developmental form of

evaluation of


Why a semantic network?Why a semantic network?

Semantic Types serve as high level categories Semantic Types serve as high level categories assigned to Metathesaurus concepts, assigned to Metathesaurus concepts, independently independently of their position in a hierarchyof their position in a hierarchy

A relationship between 2 Semantic Types (ST) is a A relationship between 2 Semantic Types (ST) is a possible link between 2 concepts that have been possible link between 2 concepts that have been assigned to those assigned to those STsSTs

The relationship may or may not hold at the concept The relationship may or may not hold at the concept levellevelOther relationships may apply at the concept levelOther relationships may apply at the concept level


Relationships can inherit semanticsRelationships can inherit semantics

Semantic Network

Metathesaurus

AdrenalCortex

AdrenalCortical

hypofunction

Disease or SyndromeBody Part, Organ,

or Organ Component

Pathologic Functionisa

Biologic Function

isa

Fully FormedAnatomical

Structure

isa

location of

location of

Aligning relationshipsAligning relationships


Relationships in the UMLSRelationships in the UMLS

MetathesaurusMetathesaurus139 relationships139 relationships

ThesauralThesauralSpecifiedSpecified

No definitionNo definitionNo organizationNo organization

Semantic NetworkSemantic Network54 relationships54 relationships

Textual definitionTextual definitionOrganized in 5 hierarchiesOrganized in 5 hierarchies

isaassociated with

functionally related tophysically related tospatially related totemporally related toconceptually related to

isacausesdue tomanifestation oficd asteriskassociated finding ofmay treat

(2005AC)


UMLS Semantic Network

UMLS Metathesaurus

Conceptcategorization

Conceptcategorization

Concept 1 Concept 2

Semantic Type aSemantic Type b

Semantic Type c

Concepts/Types vs. RelationshipsConcepts/Types vs. Relationships

Semantic Networkrelationships

Metathesaurusrelationship

?


ObjectivesObjectives

Elicit the meaning of Metathesaurus relationshipsElicit the meaning of Metathesaurus relationshipsEstablishEstablish

A correspondence betweenA correspondence betweenEach of the 139 relationships in the MetathesaurusEach of the 139 relationships in the MetathesaurusAnd relationships in the Semantic NetworkAnd relationships in the Semantic Network

The nature of the correspondenceThe nature of the correspondenceEquivalence, specialization, otherEquivalence, specialization, other

Enrich SN relationships with relationships from Enrich SN relationships with relationships from the Metathesaurusthe MetathesaurusDevelop an ontology of relationshipsDevelop an ontology of relationships


Multiple complementary approachesMultiple complementary approaches

MetathesaurusMetathesaurus--centriccentricManual elicitationManual elicitationAbstraction at the level of highAbstraction at the level of high--level conceptslevel conceptsAbstraction at the level of Semantic TypesAbstraction at the level of Semantic Types

Semantic NetworkSemantic Network--centriccentric


Manual elicitationManual elicitation

Create random samples of a maximum of 50 Create random samples of a maximum of 50 relations per Metathesaurus relationshiprelations per Metathesaurus relationshipReview manuallyReview manually

1a1a

38Lister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical CommunicationsLister Hill National Center for Biomedical CommunicationsDomain Range

causative agent ofcausative agent of


Virus Virus ………………………………………………………… Chronic aggressive viral hepatitisChronic aggressive viral hepatitisAntiAnti--psychotic agent psychotic agent ………………………… SulpirideSulpiride poisoning of undeterminedpoisoning of undetermined……Autonomic agent Autonomic agent ……………………………… LidoflazineLidoflazine allergyallergyIngestible alcoholIngestible alcohol……………………………… Mental and behavioral disordersMental and behavioral disorders

InfectiousInfectiousagentsagents

andandPharmacologicPharmacologic

substancessubstances

DisordersDisorderscausescauses

causative agent causative agent ofofSNOMEDSNOMED CTCT ≡≡ causescausesSNSN

““Brings about a condition or an effect. Implied here is Brings about a condition or an effect. Implied here is that an agent, such as for example, a pharmacologic that an agent, such as for example, a pharmacologic substance or an organism, has brought about the effectsubstance or an organism, has brought about the effect””


Compute the lowest common Compute the lowest common ancestor(sancestor(s) for ) for concepts in the domain and range of the concepts in the domain and range of the relationship, respectivelyrelationship, respectively

Abstraction at the level of highAbstraction at the level of high--level conceptslevel concepts1b1b

access instrument of access instrument of (Endoscope, Procedure)(Endoscope, Procedure)

EndoscopeEndoscope Procedure by methodProcedure by method

access instrument ofaccess instrument of


Abstraction at the level of Abstraction at the level of STsSTs

Compute the distribution of the Compute the distribution of the STsSTs for concepts for concepts in the domain and range of the relationship, in the domain and range of the relationship, respectivelyrespectively

1c1c

access instrument of access instrument of (Medical Device,(Medical Device,Diagnostic/Diagnostic/TherapTherap. or . or PrevPrev. Procedure). Procedure)


Medical DeviceMedical Device

DiagnosticDiagnosticProcedureProcedure

TherapTherap. or . or PrevPrev..ProcedureProcedure

Health CareHealth CareActivityActivity


Identifying candidates in the SNIdentifying candidates in the SN

Health CareHealth CareActivityActivity


Medical DeviceMedical DeviceDiagnostic ProcedureDiagnostic Procedure

TherapTherap. or . or PrevPrev. Procedure. Procedure

Medical DeviceMedical Device

DiagnosticDiagnosticProcedureProcedure

TherapTherap. or . or PrevPrev..ProcedureProcedure

usesuses

““Employs in the carrying out of some activity. This Employs in the carrying out of some activity. This includes applies, utilizes, employs, and avails.includes applies, utilizes, employs, and avails.””

access instrument access instrument ofofSNOMEDSNOMED CTCT ≡≡ inv(usesinv(usesSNSN))


Semantic NetworkSemantic Network--centric approachcentric approach

For a given relationship between two For a given relationship between two STsSTs in the in the SN, list all the relationships between concepts SN, list all the relationships between concepts categorized by these categorized by these STsSTs

22

ingredient ofingredient ofactive ingredient ofactive ingredient of

dose form ofdose form ofmetabolizesmetabolizes

has contraindicationhas contraindication

SubstanceSubstance Clinical DrugClinical Drugingredient ofingredient of

{ingredient of{ingredient of**, active ingredient , active ingredient ofofSNOMEDCTSNOMEDCT, dose form of, dose form of**, } , } ≡≡ ingredient ingredient ofofSNSN


Results Results MaterialsMaterials

139 specified relationships (RELA) in the 139 specified relationships (RELA) in the MetathesaurusMetathesaurusMajor contributorsMajor contributors

SNOMED CT (62)SNOMED CT (62)LOINC (15)LOINC (15)NDFNDF--RT (15)RT (15)

Most relationships are specific to one vocabularyMost relationships are specific to one vocabulary116 specific to one vocabulary116 specific to one vocabulary23 found in at least 2 vocabularies23 found in at least 2 vocabularies

(2005AC)


Results Results AlignmentAlignment

80 relationships (58%) could be aligned80 relationships (58%) could be alignedIdenticalIdentical

affectsaffects, , process ofprocess of, , ingredient ofingredient ofRoughly equivalent toRoughly equivalent to

focus focus ofofSNOMEDCTSNOMEDCT ≡≡ issue issue ininSNSN

More specific thanMore specific thanmetabolic site metabolic site ofofNDFRTNDFRT << functionally related functionally related totoSNSN

Non aligned (nonNon aligned (non--semantic relationships)semantic relationships)Lexical Lexical relatioshipsrelatioships (e.g., (e.g., britishbritish form ofform of**, , suffix suffix ofofLOINCLOINC))Mapping relationships (e.g., Mapping relationships (e.g., see see fromfromCRISPCRISP, , mapped tomapped to**))Vocabulary management (e.g., Vocabulary management (e.g., replacesreplacesSNOMEDCTSNOMEDCT))


Some issuesSome issues

Manual elicitationManual elicitationManual review by experts requiredManual review by experts required

Abstraction at the level of highAbstraction at the level of high--level conceptslevel conceptsLowest common ancestor = root of the terminology Lowest common ancestor = root of the terminology (heterogeneous set of concepts)(heterogeneous set of concepts)

Abstraction at the level of Semantic TypesAbstraction at the level of Semantic TypesMultiple Multiple STsSTs for a set of conceptsfor a set of conceptsMultiple relationships for a given pair of Multiple relationships for a given pair of STsSTs

Semantic NetworkSemantic Network--centric approachcentric approachMultiple relationships in the Metathesaurus between the Multiple relationships in the Metathesaurus between the concepts corresponding to a given pair of concepts corresponding to a given pair of STsSTs

Towards an ontology of relationshipsTowards an ontology of relationships


MotivationMotivation

Consistency checkingConsistency checkingRelations asserted between Relations asserted between STsSTsin the SN should provide constraintsin the SN should provide constraintsfor relations between Metathesaurusfor relations between Metathesaurusconcepts categorized by these concepts categorized by these STsSTs

Effective subsumption reasoningEffective subsumption reasoningRequires explicit equivalence between identical Requires explicit equivalence between identical relationships (e.g., relationships (e.g., inv(inv(duedue toto) = ) = cause ofcause of ≡≡ causescauses))Requires explicit Requires explicit subPropertysubProperty relationsrelations


Existing resourcesExisting resources

Semantic Network relationshipsSemantic Network relationships54 relationships, not formally defined, organized into a 54 relationships, not formally defined, organized into a shallow hierarchyshallow hierarchyhttp://http://semanticnetwork.nlm.nih.govsemanticnetwork.nlm.nih.gov//

GALEN relationshipsGALEN relationshipsOver 500 relationships, semiOver 500 relationships, semi--formally defined, formally defined, organized into a hierarchyorganized into a hierarchyhttp://http://www.opengalen.orgwww.opengalen.org//

OBO relation ontologyOBO relation ontology10 relationships, formally defined10 relationships, formally definedhttp://http://obofoundry.org/roobofoundry.org/ro//

http://semanticnetwork.nlm.nih.gov/

http://www.opengalen.org/

http://obofoundry.org/ro/


GALEN relations GALEN relations isIngredientOfisIngredientOf


OBO relationsOBO relations


Extending the SN relationship hierarchyExtending the SN relationship hierarchy

associated with

associated diseaseassociated with




spatiallyrelated to


pharmacokinetics of

Draftindicates *

associated with


part of

consists of

contains

connected to

interconnects

branch of

tributary of

ingredient of

spatiallyrelated to

location of

adjacent to

surrounds

traverses


co-occurs with

precedes



branch of

inv(constitutes)

part of

tributary of

direct procedure site of

indirect procedure site of

finding site of location of

procedure site of

clinically associated with sscco-occurs with

Draft

ingredient of

precise ingredient of

active ingredient of

dose form of

affects

has directmorphology

has directsubstance

associated with


spatiallyrelated to




affects

brings about

performs

occurs in

uses

manifestation of

indicates

result of

manages

treats

disrupts

complicates

interacts with

prevents

produces

causes

carries out

exhibits

practices

process of

encodes encodesgene product

has indirectmorphology

metabolicsite of

metabolizes

pathological process of

has physio-logic effect

has specimenprocedure

interprets

Draft

causative agent ofcause of

definitionalmanifestation of

exhibits

may prevent

may treat

occurs in

occurs after

treats

process of

access instrument ofinstrument of

direct device of

indirect device of

procedure device of

inv

Draft

associated with


spatiallyrelated to




degree of

analyzes

measurement of

measures

diagnoses

property of

derivative of

method of

conceptual part of

issue in

assesses effect of

developmental form of

evaluation of

analyzes

conceptual part of

component of

diagnoses may diagnose

evaluation of

measures

method ofmeasurement method of

specimen of

manifestation of *

process of

focus of

form ofdevelopmental form of

form of

contraindicated with contraindicated with

is differential diagnostic for ddx


ReferencesReferences

VizenorVizenor L, L, BodenreiderBodenreider O, Peters L, McCray AT.O, Peters L, McCray AT.Enhancing biomedical Enhancing biomedical ontologiesontologies through alignment of through alignment of semantic relationships: Exploratory approaches.semantic relationships: Exploratory approaches.AMIA AMIA AnnuAnnu SympSymp Proc 2006:804Proc 2006:804--808.808.

McCray AT, McCray AT, BodenreiderBodenreider O.O.A conceptual framework for the biomedical domain.A conceptual framework for the biomedical domain.In: Green R, Bean CA, In: Green R, Bean CA, MyaengMyaeng SH, editors. SH, editors. The semantics The semantics of relationships: an interdisciplinary perspective.of relationships: an interdisciplinary perspective.Dordrecht; Boston: Dordrecht; Boston: KluwerKluwer Academic Publishers; 2002. p. Academic Publishers; 2002. p. 181181--198. 198.


ReferencesReferences

McCray AT, Nelson SJ.McCray AT, Nelson SJ.The representation of meaning in the UMLS.The representation of meaning in the UMLS.Methods Methods InfInf Med 1995;34(1Med 1995;34(1--2):1932):193--201201

McCray AT.McCray AT.An upper level ontology for the biomedical domain.An upper level ontology for the biomedical domain.Comp Comp FunctFunct GenomGenom 2003; 4:802003; 4:80--4.4.

MedicalOntologyResearch

Olivier BodenreiderOlivier Bodenreider

Lister Hill National CenterLister Hill National Centerfor Biomedical Communicationsfor Biomedical CommunicationsBethesda, Maryland Bethesda, Maryland -- USAUSA

Contact:Contact:Web:Web:

[email protected]@nlm.nih.govmor.nlm.nih.govmor.nlm.nih.gov

Documents

Aligning relationships in the UMLS Methods and preliminary