Skip navigation

Relationships and lifecycle vocabularies

Controlled terms and semantic relations used to connect research objects, describe versions and publication stages, and represent provenance across the research lifecycle.

This resource explains how to create explicit machine-readable links between datasets, publications, software, workflows, instruments, projects, samples and other research resources.

It also provides guidance on recording versions, lifecycle dates, derivation, dependencies, replacement and provenance without confusing these distinct types of information.

ChatGPT Image 29 лип. 2026 р., 21_12_47 (4)

Resource information

Resource type: Relationship and lifecycle vocabulary guidance

Primary audience: Researchers, data stewards, repository managers, metadata specialists, research software engineers and infrastructure providers

Coverage: Citations, supplements, documentation, part–whole relations, versions, derivation, dependencies, publication stages, lifecycle dates and provenance

Primary uses: DOI metadata, repository records, research-object graphs, workflow documentation, version management and provenance exchange

Main sources: DataCite relationType, COAR Version Types, DataCite dateType and W3C PROV-O

Last reviewed: July 2026

Purpose and scope

A typed relationship states not only that two resources are connected but also how they are connected and in which direction the relationship should be interpreted.

Lifecycle metadata records the state, version and relevant dates of a resource. Provenance metadata describes the activities, inputs, outputs and agents responsible for its creation or transformation.

These forms of metadata should be coordinated but kept separate. A resource version, a publication stage, a provenance event and a related identifier represent different concepts.

Important: there is no single vocabulary that fully represents every research-object lifecycle. A practical profile normally combines DataCite relationships and dates, an appropriate version vocabulary and a provenance model.

Relationship and lifecycle families

Scholarly relationships

Connect citations, references, supplements, reviews, documentation and related publications.

Structural relationships

Represent collections, components, series, distributions and other part–whole structures.

Version and lifecycle relations

Describe editions, versions, corrections, updates, replacements and publication stages.

Provenance relationships

Connect inputs, activities, agents, methods and outputs across a research workflow.

What a typed relationship must contain

Relationship direction

DataCite relationship definitions are expressed from the registered resource A to the related resource B. The selected value must therefore be read as a complete statement beginning with the resource currently being described.

A IsDerivedFrom B: A was produced from or based on B.

A IsSourceOf B: B was produced from or based on A.

A IsNewVersionOf B: A is the newer version.

A IsPreviousVersionOf B: A is the earlier version.

A IsPartOf B: A is a component of B.

A HasPart B: B is a component of A.

DataCite relationship types

Citation and reference relationships

Supplements, descriptions and documentation

Part–whole and publication relationships

IsPartOf

The described object is a component of a collection, package, series or compound resource.

HasPart

The described object contains or aggregates the identified component.

IsPublishedIn

The resource is published inside a journal, proceedings or another publishing container.

IsVersionOf

The resource is a versioned instance rather than a structural component of the target object.

Version relationships

Derivation, dependencies and compilation

Identity, form, review and translation

Using the Other relationship

Use Other only when no existing DataCite relationship describes the connection adequately.

Supply relationTypeInformation with a concise and unambiguous explanation of the relationship.

Do not use Other merely because the correct controlled relation has not yet been identified.

A local extension should be documented and mapped to a standard relationship when a suitable value becomes available.

RelatedIdentifier and RelatedItem

RelatedIdentifier

Use when the related resource has a globally unique identifier such as a DOI, URL, IGSN, RAiD or SWHID.

RelatedItem

Use when a structured bibliographic or resource description is needed, including when no persistent identifier exists.

Relation type

Both properties use the controlled DataCite relationType vocabulary.

Local workflow links

Use a dedicated provenance representation when individual workflow steps require detailed input–output relations.

Dimensions of a resource lifecycle

Version identifier

Distinguishes successive releases or states of the resource, such as 1.0, 1.1 or 2.0.

Publication stage

Distinguishes author, submitted, accepted, proof and published versions.

Lifecycle date

Records when the resource was created, collected, issued, updated, made available or withdrawn.

Provenance event

Describes the activity that generated or transformed the resource and the responsible agents.

COAR Version Types

When to use COAR Version Types

Use COAR Version Types: for manuscripts, articles and other resources passing through a recognisable scholarly publication process.

Do not use AO, AM or VoR: as generic stages of a raw dataset, software build or computational run unless the local profile clearly adopts that publication model.

For datasets: record an explicit version number, version DOI relations and relevant lifecycle dates.

For software: use a release or semantic version identifier and connect releases through DataCite version relations.

For workflows: distinguish the workflow specification, its version and individual executions.

Dataset and software versioning

Lifecycle dates

Date type and lifecycle state

Submitted date: records when submission occurred.

Submitted version: identifies the publication stage of the resource.

Version number: distinguishes a particular release.

Version relationship: connects that release with other identified versions.

Provenance activity: explains how the version was produced.

Provenance relationships with PROV-O

PROV-O represents provenance through three principal classes: Entity, Activity and Agent.

Entity

A dataset, file, sample, model, software object, configuration or another identifiable thing.

Activity

A collection, measurement, simulation, transformation, training, validation or analysis process.

Agent

A person, organisation or software agent responsible for an entity or activity.

Qualified relation

An expanded relation carrying a role, date, plan or other contextual information.

Core provenance relationships

DataCite relationships and PROV-O

Example: DFT → MLIP → MD relationship chain

Example of DataCite relationships

{
  "doi": "10.xxxx/example-dataset-v2",
  "version": "2.0",
  "relatedIdentifiers": [
    {
      "relatedIdentifier": "10.xxxx/example-dataset",
      "relatedIdentifierType": "DOI",
      "relationType": "IsVersionOf",
      "resourceTypeGeneral": "Dataset"
    },
    {
      "relatedIdentifier": "10.xxxx/example-dataset-v1",
      "relatedIdentifierType": "DOI",
      "relationType": "IsNewVersionOf",
      "resourceTypeGeneral": "Dataset"
    },
    {
      "relatedIdentifier": "10.xxxx/example-publication",
      "relatedIdentifierType": "DOI",
      "relationType": "IsSupplementTo",
      "resourceTypeGeneral": "JournalArticle"
    },
    {
      "relatedIdentifier": "https://example.org/workflow/",
      "relatedIdentifierType": "URL",
      "relationType": "IsDocumentedBy",
      "resourceTypeGeneral": "Workflow"
    },
    {
      "relatedIdentifier": "10.xxxx/source-dataset",
      "relatedIdentifierType": "DOI",
      "relationType": "IsDerivedFrom",
      "resourceTypeGeneral": "Dataset"
    }
  ],
  "dates": [
    {
      "date": "2026-07-01",
      "dateType": "Created"
    },
    {
      "date": "2026-07-25",
      "dateType": "Issued"
    }
  ]
}

Example of workflow provenance

{
  "@context": {
    "prov": "http://www.w3.org/ns/prov#",
    "ex": "https://example.org/workflow/"
  },
  "@graph": [
    {
      "@id": "ex:reference-dataset",
      "@type": "prov:Entity",
      "prov:wasGeneratedBy": {
        "@id": "ex:dft-calculation"
      }
    },
    {
      "@id": "ex:dft-calculation",
      "@type": "prov:Activity",
      "prov:used": [
        {
          "@id": "ex:input-structures"
        },
        {
          "@id": "ex:dft-configuration"
        }
      ],
      "prov:wasAssociatedWith": {
        "@id": "ex:dft-software"
      }
    },
    {
      "@id": "ex:trained-model",
      "@type": "prov:Entity",
      "prov:wasGeneratedBy": {
        "@id": "ex:model-training"
      },
      "prov:wasDerivedFrom": {
        "@id": "ex:reference-dataset"
      }
    },
    {
      "@id": "ex:model-training",
      "@type": "prov:Activity",
      "prov:used": {
        "@id": "ex:reference-dataset"
      }
    },
    {
      "@id": "ex:md-trajectory",
      "@type": "prov:Entity",
      "prov:wasGeneratedBy": {
        "@id": "ex:md-simulation"
      }
    },
    {
      "@id": "ex:md-simulation",
      "@type": "prov:Activity",
      "prov:used": [
        {
          "@id": "ex:trained-model"
        },
        {
          "@id": "ex:initial-configuration"
        }
      ]
    }
  ]
}

How to select a relationship

1. Identify both objects

Determine exactly which resource is described and which resource is the target.

2. Form a sentence

Read the statement as “resource A [relation] resource B” before selecting the controlled value.

3. Choose the narrowest relation

Prefer a precise citation, version, derivation or part relation over a generic reference.

4. Validate the identifiers

Check the target identifier, identifier scheme, direction and resource type.

Relationship selection guide

Validation of relationships and lifecycle metadata

Direction validation

Confirm that the relation reads correctly from the described resource to the target.

Identifier validation

Check identifier syntax, type, uniqueness and resolvability where applicable.

Lifecycle validation

Check consistency between version number, version type, dates and replacement status.

Graph validation

Detect contradictory, circular or missing relationships across the research-object graph.

Recommended validation conditions

Every related identifier has an identifier type and relationship type.

The target identifier resolves to the expected resource where resolution is supported.

Inverse relations are used consistently across records where both sides are maintained locally.

A resource is not marked as both identical to and derived from the same target.

A new version does not point to itself as the previous version.

An obsolete resource points to the current replacement.

COAR publication-version terms are not applied to unrelated workflow or dataset states without documentation.

Provenance links refer to identifiable entities, activities and agents.

Common implementation errors

Providing an identifier without a relation type: the connection cannot be interpreted automatically.

Reversing relation direction: source and target acquire the opposite meaning.

Using References for every connection: citation, derivation and version semantics are lost.

Using IsPartOf for versions: structural containment is confused with version identity.

Using IsNewVersionOf for every modification: minor metadata corrections are treated as new citable resource versions.

Using IsIdenticalTo for similar objects: distinct datasets or representations are incorrectly declared identical.

Recording provenance only as narrative text: inputs, activities and outputs cannot be traversed by software.

Applying COAR publication stages to computational runs: publication lifecycle and execution lifecycle are confused.

Recording only the latest version: earlier citations and the update history cannot be resolved.

Using Other without clarification: the relationship remains semantically ambiguous.

How to use this resource

Identify the connection

Determine whether the relation concerns citation, structure, version, derivation, dependency or provenance.

Select the model

Use DataCite for stable resource links, COAR for publication versions and PROV-O for process provenance.

Record structured metadata

Preserve relation direction, target identifier, resource type, dates and version information.

Validate the graph

Check consistency across records, versions, dependencies and provenance chains.