Bridging Epoch: Mapping Two Clinical Trial Ontologies · Bridging Epoch: Mapping Two Clinical Trial...

Bridging Epoch: Mapping Two Clinical Trial Ontologies

10th International Protégé ConferenceJuly 17, 2007

Samson W. Tu1, Douglas Fridsma2, Ravi D. Shankar1, Martin O'Connor1, Amar K. Das1, David B. Parrish3

1Stanford University, USA2University of Pittsburgh, USA

3The Immune Tolerance Network, Pittsburgh, PA, USA

2

Problem: Ontologies and semantic interoperability

Ontology 1Ontology 2

3

BRIDG: Biomedical Research Integration Group

Part of US NCI Cancer Biomedical Informatics Grid (caBIG)Stakeholders include US FDA, HL7, CDISCCreate shared domain model for protocol-driven clinical research

ComprehensiveConsensus-basedAbstract and context neutral

BRIDG

4

5

6

EPOCH: Immune Tolerance Network clinical trial ontologies

Immune Tolerance Network (ITN)International collaborative research effort that sponsors clinical trials and mechanistic assays on immune tolerance

EPOCH clinical trial modelDeveloped at Stanford Medical InformaticsDesigned to provide semantic foundation for management of clinical trials

EPOCH

7

Management of clinical trials involves complex data and multiple groups

ProtocolGroup

Schedule of Events

CRO

CRF

Data Center

AccessionAccessionIDCoreID (FK)FilenameTransmissionDtTransactionCodeID (FK)RecordDtUserID (FK)

ArchiveArchiveIDCoreID (FK)ProcessID (FK)NameDirectoryRecordDtUserID (FK)

BarcodeBarcodeIDStudyID (FK)SubjectID (FK)VisitID (FK)BarcodeCollectionDtRecordDtUserID (FK)

ConfigParameterConfigParameterIDProcessID (FK)NameValueRecordDtUserID (FK)

DescriptionRecordDtUserID (FK)

Core_AssaysCoreID (FK)AssayID (FK)SpecimenTypeID (FK)RecordDtUserID (FK)

Core_ContactsCoreID (FK)ContactTypeID (FK)UserID (FK)RecordDt

Core_Issues

IssueIDCoreStudyCoresampleIDRequestedByDescriptionRequestDtResponseStatusCoreSampleID (FK)

CoreSample

CoreSampleIDStudyID (FK)AccessionID (FK)ProcessingCodeID (FK)SampleTypeID (FK)RecdDtAnalysisDtTechnameFilenameExternalIDBarcodeID9SpecimenTypeID (FK)RecordDtUserID (FK)

CoreSample_Detail

CoreSampleID (FK)BarcodeID (FK)QCSampleID (FK)ValidationCodeID (FK)ValidationDtRecordDtUserID (FK)

CoreSample_DevCodeCoreSampleID (FK)DevCodeID (FK)NotesRecordDtUserID (FK)

DevCodeDevCodeID

CodeDescriptionGetCommentsRecordDtUserID (FK)

ampleID (FK)D (FK)D(FK)FK)D(FK)(FK)meeID (FK)ntdDtD(FK)

FieldFieldIDNameDescriptionRecordDtUserID (FK)

FileRequestRequestIDRequestDtCompletedDtEmailAddressStatusFileNameFileTypeMessageArchiveID (FK)UserName

FileRequestList

FileRequestListIDStatusFileNameDirectoryRequestID (FK)Message

K)

pcrResult_Qual

CoreSampleid (FK)TestID (FK)ReplicateIDBaseVisitID (FK)QualifierAtLeastSampleQualityCalibratorQualityRecordDtUserID (FK)

ProcessProcessIDNameRecordDtUserID (FK)

ProcessingCodeProcessingCodeIDCodeDescriptionRecordDtUserID (FK)

SMS_Acc

Study_IDSite_IDParticipanVisit_NumCollectionCollectionBarcodeSpecimenLabSite_Dev_Core_DevO_Study_O_Site_IDO_ParticiO_Visit_NO_CollecO_CollecO_SpecimInvalidStaInvalidCoInvalidCoInvalidSIDInvalidPIDInvalidEnRecordDt

SMS_Issued_QueriesIssueID (FK)Query_ID (FK)RecordDtResolved

SMS_IssuesIssueIdCoreSampleID (FK)TextMsgStatusRecordDt

SMS_Qu

QueryStudyQuerySite_qQueryQueryQueryQueryDate_Date_Date_InitiatResolExternResolDate_Query

SourceSourceID

CoreID (FK)VersionDescriptionStartDtEndDtRecordDtUserID (FK)

Source_FieldsSourceFieldID

SourceID (FK)FieldID (FK)OrdinalRecordDtSkipFlagRecordDtUserID (FK)

SampleTypeSampleTypeIDNameRecordDtUserID (FK)

CodeConvNumIsLocalLabIRBApprovedlDtITNApprovedDtBudgetApprovedDtActivationDtRecordDtUserID (FK)

SubjectSubjectIDSiteID (FK)ParticipantIDStudyGroupID (FK)RecordDtUserID (FK)

TransactionCode

TransactionCodeIDCodeDescriptionRecordDtUserID (FK)

ValidationCodeValidationCodeIDCodeDescriptionRecordDtUserID (FK)

rptDataExtractRequestRequestNumberDtOfRequestDtRequiredByRequestorNameRequestorPhoneStudyNumCoreStudyDataQCSamplesResearchDevelopmentStartDtEndDtVisitsDetailDataFormatIncludeDeviationCodesIncludeTrtCohortIncludeUnvalidatedSamplesSpecialInstructionsAssignedToAssignedDtPlanStatusDtPlanAvailableDtPlanFinalRequestClosedByClosedDtSiteDetailsInformaticsNoteRecUpdatedByRecUpdatedOnDataExtractType

D(FK)reID (FK)FK)

Dt(FK)

Scee g tEnrollmentDtStudyGroupID (FK)DonorIDRecipientIDRecordDtScreeningFailureResasonIDScreeningFailureCommentUserID (FK)

ETLMappingDetailMapDtlIDSourceColumnTargetTableTargetColumnTypeMapHdrID

ETLMappingHeaderMapHdrIDMappingNameSourceVersion

GenericLoadGenLoadIDFileNameFileDirRowNumberColumnNameCellValueSourceVersiondtEnteredLoadID

GenericLoadBatchGLBatchID

FileDirRecordDtFileName

HLABusinessRuleDetailBRDetailIDTargetTableTargetColumnBRValueTargetColumnTypeBRHdrID

HLABusinessRuleHeaderBRHdrIDVersionBRName

HLADataImportLogImportLogIDFileNameStudyNumCodeVisitNumSpecimenTypeParticipantIDBarcodeDescriptionRecordDtOperatorIDAccessionID

HLAEvertLogHLAEventLogIDEventDescRecordDtOperatorIDGLBatchID

ScreeningFailureReasonScreeningFailureReasonIDCodeDescriptionEnabledStudyID

TransplantTransplantIdRecipient_SubjectIdDonor_SubjectIdOrganRecordDtOperatorStudyId

Clinical Trial Data

Cimarron

ImmunoTrak

CoreLabs

AssayResults

Tubes Manufacturer

KitReport

Assay Group

OperationsGroup

SpecimenTable

TubeTable

QueriesReports

QueriesReports

8

Three goals of EPOCHontologies

Design tools to help acquire and maintain knowledge about protocol and assay designsUse this knowledge to drive data collectionduring a trialImplement querying methods to support trial management, and ad hoc data analysis

9

EPOCH ontologies created in Protégé OWL

10

ITN wants to use BRIDG-compliant applications

ITN protocols encoded as EPOCHknowledge bases to drive caBIG applications (e.g., Patient Study Calendar)Challenge: Develop methods to

Harmonize common subset of BRIDG & EPOCH: shared semanticsOvercome representational mismatch

• Representation languages• Representation choices• (Terminological mismatch not consider here)

11

Approach taken

Semantic alignmentOvercoming representation language mismatch Overcoming representation choice mismatches

12

Approach taken

Semantic alignmentUse Excel spreadsheet to systematically review and document possible mappingsDefine necessary preconditions for mapping

Overcoming representation language mismatch Overcoming representation choice mismatches

13

Semantic Alignment: Excel spreadsheet

14

Semantic Alignment: Restrictions on EPOCH

Mapping from EPOCH to BRDG => Place restrictions on EPOCH

Only one schedule of activitiesPeriod has no subperiodsLimited temporal annotations…

Define necessary conditions for mappingFormulate as DL definition of “BRIDGClinicalTrial” subclass of epoch:ClinicalTrial

15

BRIDGClinicalTrial in EPOCH

Need trial-specific closure axioms to do automated classification of EPOCH trials that can be mapped to BRIDG

16

Approach taken

Semantic alignmentOvercoming representation language mismatch Overcoming representation choice mismatches

17

Overcoming representation language mismatch: BRIDG-in-OWL

Scope: BRIDG Study Planned View + BRIDG Complex Data Types

18

Example: PlannedStudy

19

Modifications to BRIDG

Driven by Patient Study Calendar application requirementsAdded several associational relationshipsModified some subsumption relations

20

Approach taken

Semantic alignmentOvercoming representation language mismatch Overcoming representation choice mismatches

21

Overcoming representation choice mismatch: Epoch example

Period

Screening0

Arm1Cycle1 Arm1Cycle2

PeriodType=Screening

PeriodType=Intervention PeriodType=Intervention

Arm2Cycle1 Arm2Cycle2PeriodType=Intervention PeriodType=Intervention

Intervention epochScreening epoch

EPOCH

BRIDG

22

SWRL rule to map epochs

EPOCH:periodTypes of periods correspond to BRIDG:epochsEPOCH :periodType.label corresponds toBRIDG:epoch.code.displayName

23

Successfully used an EPOCH clinical trial to configure BRIDG Patient Study Calendar application

Herold protocolin EPOCH

Herold protocolin BRIDG

Herold protocolin PSC XML

SWRL rules SWRL rules Patient StudyCalendar

Automated mappings except for onerelationship

Because of OWL/SWRL’s open-world assumption, First epoch cannot be derived as an epoch that has no predecessor

24

Conclusions

Semantic interoperability requiresHarmonization of subsets of ontologies/modelsOvercoming mismatches in representation languages and representation choices

OWL restrictions and SWRL rules help toovercome semantic and syntactic mismatchesPossible future work

Continued harmonization of BRIDG/EPOCHScalability and (semi-)automation of method

25

Thank you!

Questions?Comments?Suggestions?