HL7 Clinical Genomics – RIM Constraining Issues May 2008

HL7 Clinical Genomics –RIM Constraining Issues

May 2008

The HL7 Clinical Genomics SIG

Amnon Shabo (Shvo), PhD

HL7 Clinical Genomics SIGCo-chair and Modeling Facilitator

HL7 Structured Documents TCCDA R2 Co-editorCCD Implementation Guide Co-editor

2

Haifa Research Lab

To achieve semantic interoperability…

ClinicalTrials

Imaging

EHR

Orders& Observations

Pharmacy

ClinicalGuidelines

Health RIM

ClinicalDocuments

ClinicalGenomics

But how do we cope with the challenge of personalized healthcare?Need to bring mass genomic data into healthcare oriented standards!

…we need standard specs derived from a Central Health RIM:

Bioinformatics

Data Models

Encapsulate

& bubble-upOur s

olution

:

3

Haifa Research Lab

0..* associatedObservation

typeCode*: <= COMP

component

0..* associatedProperty

typeCode*: <= DRIV

derivedFrom2

0..* polypeptide

typeCode*: <= DRIV

derivedFrom5

SEQUENCES & PROTEOMICS

0..* expression

typeCode*: <= COMPcomponent1

0..* sequenceVariation

typeCode*: <= COMP

component3

IndividualAlleleclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]negationInd: BL [0..1]text: ED [0..1]effectiveTime: GTS [0..1]value: CD [0..1] (allele code, drawn from HUGO-HGVS or OMIM)methodCode: SET<CE> CWE [0..*]

GeneticLocusclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]code: CE CWE [0..1] (e.g., ALLELIC, NON_ALLELIC)text: ED [0..1]effectiveTime: IVL<TS> [0..1]confidentialityCode: SET<CE> CWE [0..*] <= ConfidentialityuncertaintyCode: CE CNE [0..1] <= Uncertaintyvalue: CD [0..1] (identifying a gene through GenBank GeneID with an optional translation to HUGO name.)methodCode: SET<CE> CWE [0..*]

0..* individualAllele

typeCode*: <= COMP

component1

SequenceclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]code: CD CWE [1..1] (the sequence standard code, e.g. BSML)text: ED [0..1] (sequence's annotations)effectiveTime: GTS [0..1]uncertaintyCode: CE CNE [0..1] <= Uncertaintyvalue: ED [1..1] (the actual sequence)interpretationCode: SET<CE> CWE [0..*] <= ObservationInterpretationmethodCode: SET<CE> CWE [0..*] (the sequencing method)

ExpressionclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]code: CE CWE [1..1] (the standard's code (e.g., MAGE-ML identifier)negationInd: BL [0..1]text: ED [0..1]effectiveTime: GTS [0..1]uncertaintyCode: CE CNE [0..1] <= Uncertaintyvalue: ED [1..1] (the actual gene or protein expression levels)interpretationCode: SET<CE> CWE [0..*] <= ObservationInterpretationmethodCode: SET<CE> CWE [0..*]

PolypeptideclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]text: ED [0..1]effectiveTime: GTS [0..1]value: CD [0..1] (protein code, drawn from SwissProt, PDB, PIR,HUPO, etc.)methodCode: SET<CE> CWE [0..*]

DeterminantPeptidesclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]text: ED [0..1]effectiveTime: GTS [0..1]value: CD [0..1] (peptide code, drawn from referencedatabases like those used in the Polypeptide class)methodCode: SET<CE> CWE [0..*]

Constrained to a restrictedMAGE-ML constrained schema,specified separately.

Constraint: GeneExpression.value

Note:A related allele that is ona different locus, and hasinterrelation with thesource allele, e.g.,translocated duplicatesof the gene.

0..* clinicalPhenotype

typeCode*: <= PERTpertinentInformation

ExternalObservedClinicalPhenotypeclassCode*: <= OBSmoodCode*: <= EVNid*: II [1..1] (The unique id of an external observation residing outside of the instance)code: CD CWE [0..1]text: ED [0..1]effectiveTime: GTS [0..1]

Note:An external observation is preferably a valid observationinstance existing in any other HL7-compliant instance,e.g., a document or a message.Use the id attribute of this class to point to the uniqueinstance identifier of that observation.

Note:A phenotype which has been actuallyobserved in the patient representedinternally in this model.

Note:This is a computed outcome, i.e.,the lab does not test for the actualprotein, but secondary processespopulate this class with thetranslational protein.

SequenceVariationclassCode*: <= OBSmoodCode*: <= EVNid: II [0..1]code: CD CWE [0..1]negationInd: BL [0..1]text: ED [0..1]effectiveTime: GTS [0..1]value: ANY [0..1] (The variation itself expressed with recognized notation like 269T>C or markup like BSML or drawn from an external reference like OMIM or dbSNP.)interpretationCode: SET<CE> CWE [0..*] <= ObservationInterpretationmethodCode: SET<CE> CWE [0..*]

KnownClinicalPhenotypeclassCode*: <= OBSmoodCode*: <= DEFcode: CD CWE [0..1]text: ED [0..1]effectiveTime: GTS [0..1]uncertaintyCode: CE CNE [0..1] <= ActUncertaintyvalue: ANY [0..1]

Note:These phenotypes are not the actual (observed)phenotypes for the patient, rather they are thescientifically known phenotypes of the sourcegenomic observation (e.g., known risks of amutation or know responsiveness to a medication).

Note:Code: COPY_NUMBER, ZYGOSITY, DOMINANCY, GENE_FAMILY,etc. For example, if code = COPY_NUMBER, then the value is oftype INT and is holding the no. of copies of this gene or allele.


typeCode*: <= PERT

pertinentInformation

EXPRESSION DATA

SEQUENCE VARIATIONS

Polypeptide

Note:The Expression class refers to both gene and proteinexpression levels. It is an encapsulating class that allowsthe encapsulation of raw expression data in its value attribute.

0..* sequence

typeCode*: <= COMPcomponent2

0..* clinicalPhenotypetypeCode*: <= PERT



typeCode*: <= PERT


Note:The code attribute indicates inwhat molecule the variation occurs,i.e., DNA, RNA or Protein.

0..* expression

typeCode*: <= COMP

component5

Note:Use the associations to the shadowclasses when the data set type (e.g.,expression) is not at deeper levels(e.g., allelic level) and needs to beassociated directly with the locus(e.g., the expression level is thetranslational result of both alleles).

0..* associatedObservationtypeCode*: <= COMP

component2

0..1 associatedObservation

typeCode*: <= COMP

component4 Note:This recursive associationenables the association of anRNA sequence derived froma DNA sequence and apolypeptide sequence derivedfrom the RNA sequence.


typeCode*: <= PERT


Note:

This class is a placeholder for a specific locus on the genome - that is - a position of a particulargiven sequence in the subject’s genome or linkage map.Note that the semantics of the locus (e.g., gene, marker, variation, etc.) is defined by data assignedin the code & value attributes of this class, and also by placing additional data relating to thislocus into the classes associated with this class like Sequence, Expression, etc..

Note:The term 'Individual Allele' doesn't refer necessarily to aknown variant of the gene/locus, rather it refers to theindividual patient data regarding the gene/locus and mightwell contain personal variations w/unknown significance.

AssociatedObservationclassCode*: <= OBSmoodCode*: <= EVNid: SET<II> [0..*]code: CD CWE [0..1]text: ED [0..1]effectiveTime: GTS [0..1]value: ANY [0..1]methodCode: SET<CE> CWE [0..*]

Note:The code attribute could hold codes likeNORMALIZED_INTENSITY, P_VALUE, etc.The value attribute is populated based on theselected code and its data type is then setupaccordingly during instance creation.

Note:The code attribute could hold codes like TYPE,POSITION.GENOME, LENGTH, REFERENCE, REGION, etc..The value attribute is populated based on the selected codeand its data type is then setup accordingly during instancecreation. Here are a few examples:If code = TYPE, then the value is of type CV and holds one of thefollowing: SNP (tagSNP), INSERTION, DELETION,TRANSLOCATION, etc.

if code = POSITION, then value is of type INT and holdsthe actual numeric value representing the variation positionalong the gene.

if code = LENGTH, then value is of type INT and holdsthe actual numeric value representing the variation length.

If code = POSITION.GENE, then value is of type CV and is oneof the following codes:INTRON, EXON, UTR, PROMOTER, etc.

If code = POSITION.GENOME, then value is of type CV and is oneof the following codes:NORMAL_LOCUS, ECTOPIC, TRANSLOCATION, etc.

If the code = REFERENCE, then value istype CD and holds the reference gene identifier drawn from areference database like GenBank.

The full description of the allowed vocabularies for codes and itsrespective values could be found in the specification.

AssociatedObservation

Note:Code: CLASSIFICATION, etc.For example, if code =CLASSIFICATION, then the valueis of type CV and is holding eitherKNOWN or NOVEL.

reference

0..* geneticLocus

typeCode*: <= REFR

Note:A related gene that is on adifferent locus, and stillhas significant interrelationwith the source gene (similarto the recursive associationof an IndividualAllele).

ClinicalPhenotypeclassCode*: <= ORGANIZERmoodCode*: <= EVN

0..* observedClinicalPhenotype

typeCode*: <= COMP

component1

0..* knownClinicalPhenotype

typeCode*: <= COMP

component2

0..* externalObservedClinicalPhenotype

typeCode*: <= COMP

component3

At least one of the target acts ofthe three component act relationshipsshould be populated, since this isjust a wrapper class.

Constraint: ClinicalPhenotype

Note:- code should indicate the type of source, e.g., OMIM- text could contain pieces from research papers- value could contain a phenotype code if known (e.g., if it’s a disease, then the disease code)

ClinicalPhenotype

ClinicalPhenotype

ClinicalPhenotype

ClinicalPhenotype

ClinicalPhenotype

ClinicalPhenotype

0..1 identifiedEntity

typeCode*: <= SBJcontextControlCode: CS CNE [0..1] <= ContextControl "OP"

subject

reference

0..* individualAllele

typeCode*: <= REFR

ObservedClinicalPhenotype

Note:This CMET might be replacedwith the Clinical Statement SharedModel for richer expressivity, whenthe that mode is approved(currently in ballot).

Constrained to a restricted BSMLcontent model, specified in aseparate schema.

Constraint: Sequence.value

0..* sequence

typeCode*: <= COMP

component4

0..* sequenceVariation

typeCode*: <= COMP

component3

AssociatedPropertyclassCode*: <= OBSmoodCode*: <= EVNcode: CD CWE [0..1]text: ED [0..1]value: ANY [0..1]


typeCode*: <= DRIVderivedFrom1

AssociatedObservation

0..* associatedObservation

typeCode*: <= COMP

component

AssociatedPropertyAssociatedObservation


typeCode*: <= DRIV

derivedFrom

AssociatedProperty0..* associatedProperty


AssociatedObservation0..* associatedObservation

typeCode*: <= COMPcomponent

0..* sequenceVariationtypeCode*: <= DRIV

derivedFrom3derivedFrom2

0..* sequence

typeCode*: <= DRIV

0..* determinantPeptides

typeCode*: <= DRIV

derivedFrom4

0..* determinantPeptides

typeCode*: <= DRIVderivedFrom


typeCode*: <= PERT

pertinentInformation 0..* clinicalPhenotype

typeCode*: <= PERT


AssociatedProperty


typeCode*: <= DRIV

derivedFrom

AssociatedProperty

GeneticLociclassCode*: <= OBSmoodCode*: <= EVNid: SET<II> [0..*]code: CD CWE [0..1]effectiveTime: GTS [0..1]value: ANY [0..1]

0..* geneticLocitypeCode*: <= COMPcomponentOf


typeCode*: <= PERTpertinentInformation

GeneticLoci

0..* geneticLoci

typeCode*: <= COMP

componentOf

GeneticLoci

0..* geneticLoci

typeCode*: <= COMP

componentOf

0..* polypeptide


Polypeptide

0..* polypeptide

typeCode*: <= DRIV

derivedFrom2

Note:Use this class to indicate a set of genetic locito which this locus belongs. The loci set couldbe a haplotype, a genetic profile and so forth.Use the id attribute to point to the GeneticLociinstance if available. The other attributesserve as a minimal data set about the loci group.

PHENOTYPES

Note:Any observation related to the variation and is notan inherent part of the variation observation (the lattershould be represented in the AssociatedProperty class).For example, the zygosity of the variation.

Note:Use this class to point to a variationgroup to which this variation belongs.For example, a SNP haplotype.

Note:Any observation related to the sequence and is notan inherent part of the sequence observation (the lattershould be represented in the AssociatedProperty class).For example, splicing alternatives.

Note:Key peptides in the proteinthat determine its function.

Note:There could be zero to manyIndividualAllele objects in aspecific instance. A typicalcase would be an allele pair,one on the paternalchromosome and one on thematernal chromosome.

Note:Use this class toshow an allelehaplotype like in HLA.

Note:Any observationrelated to theexpression assayand is not aninherent part ofthe expressionobservation.

Note:Use this class forinherent dataabout the locus, e.g.chromosome no.

IdentifiedEntityclassCode*: <= IDENTid: SET<II> [0..*]code: CE CWE [0..1] <= RoleCode

Note:Use this role to identify a different subject(e.g., healthy tissue, virus, etc.) than theone propagated from the wrappingmessage or payload (e.g., GeneticLoci).

ScopingEntityclassCode*: <= LIVdeterminerCode*: <= INSTANCEid: SET<II> [0..*]code: CE CWE [0..1] <= EntityCode

0..* assignedEntity

typeCode*: <= PRFcontextControlCode: CS CNE [0..1] <= ContextControl "OP"

performer

0..*

performer

0..*

performer1

0..*

performer2

0..*

performer1

0..*

performer2

Genetic Locus(POCG_RM000010)

The entry point tothe GeneticLocus modelis any locus on the genome.

Constrained to a restricted MAGE-MLcontent model, specified in aseparate schema.

Constraint: Expression.value

Expression

Sequence

SequenceVariation

SequenceVariation

0..* clinicalPhenotypetypeCode*: <= PERT


ClinicalPhenotype

CMET: (ASSIGNED) R_AssignedEntity

[universal](COCT_MT090000)

0..1 scopedRoleName

CMET: (ACT) A_SupportingClinicalInformation

[universal](COCT_MT200000)

The Locus and its Alleles

SequenceVariations

ExpressionData

Sequenceand

Proteomics

ClinicalPhenotypes

The HL7 v3 DSTU GeneticLocus Model - Focal Areas:

4

Haifa Research Lab

The Underlying Paradigm: Encapsulate & Bubble-up

End user Applications

for clinical practice & research

Genomic Data Sources

EHR System

HL7 CG Messages with m

ainly

Encapsulating HL7 Objects HL7 C

G M

essa

ges

with

enca

psul

ated

dat

a as

soci

ated

with

HL7 c

linic

al o

bjec

ts (p

heno

type

s)

Bubble up the most clinically-significant raw

genomic data into specialized HL7 objects and

link them with clinical data from the patient EHR

Decision Support Applications

Knowledge(KBs, Ontologies, registries,

reference DBs, Papers, etc.)

Bridging is the challenge…

Encapsulation by predefined & constrained

bioinformatics schemas

Bubbling-up is done continuously

by specialized Decision Support

applications

5

Haifa Research Lab

The Refinement Process Starts in the RIM

Observation

Act Specialization

DiagnosticImage

Observation SpecializationPublicHealthCase

Observation Specialization

So why not a genomic specialization?

Where do we draw the line and stop specializing?

The core classes:An Entity plays a Role which Participates in an Act

Generic ontology

6

Haifa Research Lab

Refining a RIM Class in a Static Model

This process of refinement is needed in Clinical Genomics Clinical Genomics brings new & unique concepts into HL7 Therefore we proposed new RIM classes but got rejected

e.g., SequenceVariance with attributes like position, length, etc.

We then developed the GEN code hierarchy instead…

classCode attr.

Clone B. Name

code attr.

OBS

BaseBattery

LOINC code for

Test Battery/Panel

OBS

BaseUnitaryResult

LOINC code for

Test

Examples from the RCRIM CT Lab Model:

OBS

ToxicityGrade

LOINC

7

Haifa Research Lab

The GEN Hierarchy (in ActClass vocabulary)

8

Haifa Research Lab

The GEN Hierarchy (in ActClass vocabulary)

The GEN code hierarchy was accepted by RIM Harmonization already in July 2006!

It indentifies core classes of the genomic models Enables further specialization of classes like sequence variation

by using the code attribute

It allows CDSS to implement the bubble-up paradigm, i.e., creating and looking for genotypephenotype associations

We need minor refinements of this hierarchy, e.g.: Add ALLELE code to the GEN hierarchy This is also addressing Bob Dolin’s ballot comment about SEQVAR for

IndividualAllele wild type which is not an appropriate code PHN (Phenotype) is a genomic-related observation (edit GEN definition)

9

Haifa Research Lab

Complex GenotypePhenotype Associations

“Some patients, particularly young children, may be born with a genetic mutation which means they are at risk of hearing loss after taken antibiotics called aminoglycosides. There is now a drive to consider screening patients for the genetic mutation known as m.1555A-G which is held in around 1 in 1,611 newborns in the USA, 1 in 206 newborns in New Zealand and 1 in 40,000 newborns in the UK. Aminoglycosides are valuable antibiotics used for serious infections such as complicated urinary tract infections, TB and septicemia. They are known to potentially cause damage to the ear - otoxicity. Individuals holding the mutation have an inherited predisposition which makes them extremely sensitive to the effects - they can end up with severe and permanent hearing loss."

10

Haifa Research Lab

The PHN (Phenotype) Class Code

Signifies that an observation is a phenotype

Typically is associated with a source genomic observation But not necessarily in cases where the exact genetic sources is yet unknown

A genomic observation could be associated with other genomic observations or with other observations not necessarily classified as phenotypes

Prepare the ground for a ‘disease model’ expressed in RIM so that patient data could be better checked against it

The IBM Clinical Genomics solution as used the Hypergenes project - http://www.hypergenes.eu/) relies on the GEN hierarchy and performs semantic computations based on these codes

11

Haifa Research Lab

General Vocabulary Issues

What are the clear-cut criteria that distinguish between ActClass and ActCode, preferably in a way that a designer of a clinical decision support application could use them programmatically?

ActCode should specialize ActClass but current guidance and the actual values in the two vocabularies sometime blur the differences between the two, e.g.: _MedicationObservationType act code doesn’t specialize any

act class SpecimenTransport act code could be specializing TRANS or

SPCTRT _ActCognitiveProfessionalServiceCode act code doesn’t

specialize a.class

If indeed one specializes the other, why not merge them into one coherent vocabulary where sub-typing will become explicit? The fields classCode, [clone name] and code will bind to the

appropriate layer in this proposed consolidated vocabulary

12

Haifa Research Lab

What if GEN is Moved to ActCode…

In lack of class code & clone name - need to rely on the code attribute, however…

This poses the following problems: Code attribute is often drawn from external terminology

Giving up class code & clone name leads to loss of granularity in the data represented by the class

Inferring the generic GEN code from a value in the code attribute might not be easy, especially if drawn from external terminology

13

Haifa Research Lab

Clone Business Names DO Carry Semantics

Current guidance: Clone business names should not carry semantics

Rebuttal: These names are meaningful to committee members and are balloted and approved by the HL7 membership

An example – CDA AuthoringDevice vs. Device The CDA R2 AuthoringDevice and Device clones have the same class code

(DEV) and the same binding of the code attribute (EntityCode). In fact, they have the very same set of attributes refined exactly the same

(cardinality, coding strength, vocab, etc). The only way to distinguish between the two is through the clone name

(unless you rely on the traversal path). The two clones are semantically distinct: AuthoringDevice has to do with the

authoring of the document and Device has to do with some specimen mentioned in the body of the document.

In addition, the clone AuthoringDevice does carry semantics which cannot be found in either EntityClass or EntityCode vocabularies.

DeviceclassCode*: <= DEVdeterminerCode*: <= INSTANCEcode: CE CWE [0..1] <= EntityCodemanufacturerModelName: SC CWE [0..1] <= ManufacturerModelNamesoftwareName: SC CWE [0..1] <= SoftwareName

AuthoringDeviceclassCode*: <= DEVdeterminerCode*: <= INSTANCEcode: CE CWE [0..1] <= EntityCodemanufacturerModelName: SC CWE [0..1] <= ManufacturerModelNamesoftwareName: SC CWE [0..1] <= SoftwareName

???

14

Haifa Research Lab

Clone Business Names (cont.)

The CDA R2 Encounter and EncompassingEncounter have the same class code (ENC) and the same binding of the code attribute (ActEncounterCode). Semantically they're distinct: the EncompassingEncounter is the encounter documented in the CDA instance and the other clone represents an encounter you may refer to in the body of the document.

A different example of the clone business name semantics is from Patient Administration (Emergency Encounter) model: the clone ValuablesLocation has class code<=OBS, no code attribute and value attribute that is not bound to any vocabulary (it also has mood & negation attributes). Obviously, the semantics carried by ValuablesLocation cannot be found under OBS but also cannot be found in both ActClass and ActCode vocabs.

EncounterclassCode*: <= ENCmoodCode*: <= x_DocumentEncounterMoodid: SET<II> [0..*]code: CD CWE [0..1] <= ActEncounterCodetext: ED [0..1]statusCode: CS CNE [0..1] <= ActStatuseffectiveTime: IVL<TS> [0..1]priorityCode: CE CWE [0..1] <= ActPriority

EncompassingEncounterclassCode*: <= ENCmoodCode*: <= EVNid: SET<II> [0..*]code: CE CWE [0..1] <= ActEncounterCodeeffectiveTime*: IVL<TS> [1..1]dischargeDispositionCode: CE CWE [0..1] <= EncounterDischargeDisposition

???

15

Haifa Research Lab

Formalize Clone Business Names…

Formalize the clone business name into a vocabulary domain, intertwined in the consolidated ActClass/Code

Adding a definition for each clone name is already done in the walk-through of each model, and possibly in the glossary

In this way, we can relate to clone names as codes in the cascading 'identification' of a class

This could ease the burden of class identification by the class code attribute, if indeed the wish is to keep the ActClass as minimal as possible

16

Haifa Research Lab

Thank You for Your Attention…

Questions? Comments: [email protected]

HL7 Clinical Genomics – RIM Constraining Issues May 2008

Documents

central health rim

healthcare oriented

haifa research labthe

standard specs

mass genomic data

invariant x