Class AnnotationRelationDao.RelationSummary

java.lang.Object

ubic.gemma.persistence.service.common.description.AnnotationRelationDao.RelationSummary

Enclosing interface:

AnnotationRelationDao

public static class AnnotationRelationDao.RelationSummary
extends Object

A relation aggregated over its attesting experiments, for one basis.

One summary per (triple, basis), and deliberately not merged across bases. Merging would require deciding that two bases naming different terms mean the same thing, and they routinely do while naming different terms: MONDO's germline axiom for SURF1 points at MONDO:0700250 (mitochondrial complex IV deficiency, nuclear type 1) where the curator wrote MONDO:0009723 (Leigh syndrome). Those two share no xref, sit in different branches, and neither subsumes the other - the ontology is modelling the molecular diagnosis and the curator the clinical syndrome. Both are correct. A merge keyed on term identity would call that a disagreement; a merge keyed on anything looser would invent an equivalence nobody asserted. So the rows come back side by side and the caller sees both framings.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	AnnotationRelationDao.RelationSummary.RelationSummaryBuilder

Constructor Summary

Constructors

Constructor

Description

RelationSummary(String subjectValue, String subjectValueUri, String subjectCategory, String subjectCategoryUri, String predicate, String predicateUri, String objectValue, String objectValueUri, String objectCategory, String objectCategoryUri, Long taxonId, String taxonCommonName, Integer taxonNcbiId, AnnotationRelationBasis basis, AnnotationRelationStatus status, String source, String sourceVersion, long numberOfExperiments, long numberOfExperimentsAtFactorValue, long numberOfExperimentsAtTag, long numberOfExperimentsAtBioMaterial, Long exampleExperimentId, long numberOfExperimentsWithSubject, long objectBreadth, long subjectBreadth, String evidence)

🛑 Build one with builder(), not with this. Twenty-six positional arguments, a dozen of them long and most of the rest String, is a constructor where a transposition compiles and passes — and adding a field to it broke a stale test class three times in one day, because Maven does not recompile a test whose source has not changed and the mismatch therefore arrives at runtime as NoSuchMethodError rather than at compile time.

Method Summary

Modifier and Type	Method	Description
static AnnotationRelationDao.RelationSummary.RelationSummaryBuilder	builder()
AnnotationRelationBasis	getBasis()
String	getEvidence()
Long	getExampleExperimentId()
String	getImpliedObject()
String	getImpliedObjectUri()
String	getImpliedPredicate()
String	getImpliedPredicateUri()
String	getImpliedSubject()	The claim this relation licenses, as a triple in its own right — not the stored row read backwards.
String	getImpliedSubjectUri()
String	getImpliedTripleKey()	Identity of the CLAIM this row licenses, as opposed to getTripleKey(), which is the identity of the row as stored.
RelationInferenceDirection	getInferenceDirection()	Which end of this relation licenses an inference about the other.
long	getNumberOfExperiments()
long	getNumberOfExperimentsAtBioMaterial()
long	getNumberOfExperimentsAtFactorValue()
long	getNumberOfExperimentsAtTag()
long	getNumberOfExperimentsWithSubject()
long	getObjectBreadth()
String	getObjectCategory()
String	getObjectCategoryUri()
String	getObjectValue()
String	getObjectValueUri()
String	getPredicate()
String	getPredicateUri()
double	getScore()	Ranking score, highest first: the basis rank dominates, and support x specificity orders within a basis.
String	getSource()
String	getSourceVersion()
double	getSpecificity()	The fraction of the subject's experiments this object accounts for, or 0 when there is no denominator to divide by.
AnnotationRelationStatus	getStatus()
long	getSubjectBreadth()
String	getSubjectCategory()
String	getSubjectCategoryUri()
String	getSubjectValue()
String	getSubjectValueUri()
String	getTaxonCommonName()
Long	getTaxonId()
Integer	getTaxonNcbiId()
RelationTopicality	getTopicality()	Whether this relation says something about the subject TERM, or records a parameter of one experiment.
String	getTripleKey()

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

RelationSummary

public RelationSummary(String subjectValue,
 @Nullable
 String subjectValueUri,
 @Nullable
 String subjectCategory,
 @Nullable
 String subjectCategoryUri,
 @Nullable
 String predicate,
 @Nullable
 String predicateUri,
 String objectValue,
 @Nullable
 String objectValueUri,
 @Nullable
 String objectCategory,
 @Nullable
 String objectCategoryUri,
 @Nullable
 Long taxonId,
 @Nullable
 String taxonCommonName,
 @Nullable
 Integer taxonNcbiId,
 AnnotationRelationBasis basis,
 AnnotationRelationStatus status,
 @Nullable
 String source,
 @Nullable
 String sourceVersion,
 long numberOfExperiments,
 long numberOfExperimentsAtFactorValue,
 long numberOfExperimentsAtTag,
 long numberOfExperimentsAtBioMaterial,
 @Nullable
 Long exampleExperimentId,
 long numberOfExperimentsWithSubject,
 long objectBreadth,
 long subjectBreadth,
 @Nullable
 String evidence)

🛑 Build one with builder(), not with this. Twenty-six positional arguments, a dozen of them long and most of the rest String, is a constructor where a transposition compiles and passes — and adding a field to it broke a stale test class three times in one day, because Maven does not recompile a test whose source has not changed and the mismatch therefore arrives at runtime as NoSuchMethodError rather than at compile time.

With a builder an added field is additive: existing callers keep compiling AND keep meaning what they meant, and the new field simply defaults.

Method Details

getSpecificity

public double getSpecificity()

The fraction of the subject's experiments this object accounts for, or 0 when there is no denominator to divide by.

This is what separates a real relation from a background co-occurrence, and support on its own cannot. Abca4 null is annotated retinal degeneration nearly every time it appears; C57BL/6J appears against obesity a handful of times out of many hundreds of experiments, and against hundreds of other diseases besides. Ranking on raw support puts the background strain on top of every disease in the corpus - the obesity is diet-induced, the stroke surgical, the Burkitt lymphoma belongs to the cell line. The same measure demotes the drug-versus-model confusion with no special case, since a drug tested against everything has low specificity against any one thing.

getScore

public double getScore()

Ranking score, highest first: the basis rank dominates, and support x specificity orders within a basis.

The basis dominating is the point - an assertion outranks any amount of co-occurrence. A score alone is never enough to act on, which is why the basis, the counts and the denominator all stay on the row for a client to apply its own bar to.

getSubjectValue

public String getSubjectValue()

getSubjectValueUri

@Nullable
public String getSubjectValueUri()

getSubjectCategory

@Nullable
public String getSubjectCategory()

getSubjectCategoryUri

@Nullable
public String getSubjectCategoryUri()

getPredicate

@Nullable
public String getPredicate()

getPredicateUri

@Nullable
public String getPredicateUri()

getObjectValue

public String getObjectValue()

getObjectValueUri

@Nullable
public String getObjectValueUri()

getObjectCategory

@Nullable
public String getObjectCategory()

getObjectCategoryUri

@Nullable
public String getObjectCategoryUri()

getTaxonId

@Nullable
public Long getTaxonId()

getTaxonCommonName

@Nullable
public String getTaxonCommonName()

getTaxonNcbiId

@Nullable
public Integer getTaxonNcbiId()

getBasis

public AnnotationRelationBasis getBasis()

getStatus

public AnnotationRelationStatus getStatus()

getSource

@Nullable
public String getSource()

getSourceVersion

@Nullable
public String getSourceVersion()

getNumberOfExperiments

public long getNumberOfExperiments()

getNumberOfExperimentsAtFactorValue

public long getNumberOfExperimentsAtFactorValue()

getNumberOfExperimentsAtTag

public long getNumberOfExperimentsAtTag()

getNumberOfExperimentsAtBioMaterial

public long getNumberOfExperimentsAtBioMaterial()

getExampleExperimentId

@Nullable
public Long getExampleExperimentId()

getNumberOfExperimentsWithSubject

public long getNumberOfExperimentsWithSubject()

getObjectBreadth

public long getObjectBreadth()

getSubjectBreadth

public long getSubjectBreadth()

See Also:

• subjectBreadth

getEvidence

@Nullable
public String getEvidence()

See Also:

• evidence

getTopicality

public RelationTopicality getTopicality()

Whether this relation says something about the subject TERM, or records a parameter of one experiment. See RelationTopicality.

getInferenceDirection

public RelationInferenceDirection getInferenceDirection()

Which end of this relation licenses an inference about the other. A relation is readable from both ends and inferable from only one — see RelationInferenceDirection.

getImpliedSubject

@Nullable
public String getImpliedSubject()

The claim this relation licenses, as a triple in its own right — not the stored row read backwards.

Gemma stores Alzheimer disease --has_genotype--> APP/PS1, but the fact that follows from it is "APP/PS1 is a disease model of Alzheimer's": different ends, and a different verb. A consumer handed only the stored row has to invert it and choose a verb itself, and three consumers will choose three.

Taxon picks the verb, which is the rule the design settled on and the reason RO_0003301 has role in modeling was added to the vocabulary. A mouse carrying APP/PS1 models Alzheimer disease; a human line carrying LRRK2 G2019S is not modelling Parkinson disease, it has it. Unknown taxon falls to has role in modeling, the weaker of the two claims, rather than being guessed at.

Null when getInferenceDirection() is RelationInferenceDirection.NEITHER — nothing is implied, so there is no claim to phrase.

getImpliedSubjectUri

@Nullable
public String getImpliedSubjectUri()

getImpliedObject

@Nullable
public String getImpliedObject()

getImpliedObjectUri

@Nullable
public String getImpliedObjectUri()

getImpliedPredicateUri

@Nullable
public String getImpliedPredicateUri()

See Also:

• getImpliedSubject()

getImpliedPredicate

@Nullable
public String getImpliedPredicate()

getImpliedTripleKey

@Nullable
public String getImpliedTripleKey()

Identity of the CLAIM this row licenses, as opposed to getTripleKey(), which is the identity of the row as stored. Null when the row licenses nothing.

🛑 Two different stored relations routinely derive one claim, and tripleKey cannot see it. uib measured it on the BRCA1 card, 2026-08-18:

BRCA1         --has disease-->  breast cancer   tripleKey  …/672 RO_0016002 …/MONDO_0007254
breast cancer --has_genotype--> BRCA1           tripleKey  …/MONDO_0007254 GENO_0000222 …/672

Two keys, one claim — BRCA1 — has disease → breast cancer — rendered twice, word for word. tripleKey groups the side-by-side rows ONE relation produces across bases, which is a different shape from two relations converging on one claim, so a consumer reaching for it to deduplicate a card gets a duplicate anyway. Dedup belongs at the level the reader actually sees, which is the claim.

It correctly keeps apart claims whose verb differs: the same genotype relation on a mouse implies has role in modeling and on a human line has disease, and those are two claims rather than one rendered twice.

getTripleKey

public String getTripleKey()

builder

public static AnnotationRelationDao.RelationSummary.RelationSummaryBuilder builder()