Vocabulary
The graph uses three controlled vocabularies to structure every node and connection: entity types, research domains, and relationship predicates. This page documents each one. For contribution guidelines, editorial standards, and the inclusion criteria for new nodes, see CONTRIBUTING.md.
Entity types
Each node carries a type/ tag that determines its directory, its shape in the graph view, and the role it plays in the ecosystem. Each tag is a clickable link that opens the index of all nodes carrying that type. A node may carry a second type tag when it genuinely operates in two distinct roles.
Actors
Entities with institutional identity: staff, governance, and physical or organisational presence.
| Type | Description |
|---|---|
| institute | A research-performing organisation: an institute, university, or national research body with its own staff and governance. |
| consortium | A multi-institution collaboration that jointly produces research data or findings. The consortium itself conducts the research. |
| programme | A time-limited funded activity organised around a defined research or infrastructure objective, typically with a funding period and deliverables. |
| biobank | A large-scale biological or clinical collection from human participants, providing samples or data for research reuse. |
| facility | A physical or compute service provider (from a single-institution core lab to a national facility network) that researchers use directly for imaging, sequencing, or computing. |
Standards
Formal specifications defining how data is encoded, organised, or described.
| Type | Description |
|---|---|
| fileformat | A specification defining how data is encoded at the file level: the byte structure or syntax of a file on disk. Examples: NIfTI, FASTQ, EDF, VCF. |
| ontology | A controlled vocabulary used to annotate entities, describing what something is, what it does, or where it is in the body. Examples: HPO, SNOMED CT, UBERON. |
| datamodel | A specification defining how data is organised or related: folder structures, schemas, or relational models. BIDS and NWB are data models. They specify organisation, not file encoding. |
| protocol | A formal procedure standard for research conduct, data collection, or reporting. Example: Open Brain Consent. |
Resources
Digital systems: data repositories and software tools with no independent institutional identity.
| Type | Description |
|---|---|
| repository | A digital archive where research outputs are deposited by contributors and retrieved by users. Examples: OpenNeuro, EGA, Zenodo. |
| tool | A software system or digital service for managing, processing, discovering, or accessing research data, but not primarily a data archive. Examples: LORIS, REDCap, DataLad, FAIRsharing. |
Governance
Entities whose primary activity is coordinating, regulating, or setting norms for the open science ecosystem.
| Type | Description |
|---|---|
| authority | An independent governmental or supranational regulatory body with legal mandate to issue binding decisions, authorisations, or enforcement actions. Covers national data protection authorities, medicines regulators, and supranational supervisory bodies. Examples: CNIL, EDPB. |
| policy | A legal instrument, institutional mandate, or funder requirement with enforcement conditions. Examples: GDPR, ANR Open Science Policy, EHDS. |
| principles | A voluntary normative framework that informs policy without itself carrying enforcement. Examples: FAIR Principles, DORA, CoARA, ECoC. |
| network | A coordinating or federating body that connects institutions or national nodes into a shared infrastructure. Distinguished from a consortium (which produces research) and a working group (which produces specifications). Examples: EOSC, ELIXIR, BBMRI-ERIC, NFDI. |
| working-group | A community body that produces specifications, governance frameworks, or endorsement decisions. Examples: BIDS Steering Group, GA4GH, INCF, RDA. |
Research domains
Each node also carries domain/ tags. These specify the research modalities the node is directly involved in. A node can carry several when it spans more than one. Nodes with a general science function, such as infrastructure, cross-domain governance, and broad policy, have no domain tag. Clicking any tag below opens all nodes in that domain, so the tags double as the entry point to each modality’s slice of the graph.
| Domain | Covers |
|---|---|
| neuroimaging | MRI, fMRI, PET, and related structural or functional brain imaging |
| electrophysiology | EEG, MEG, iEEG, LFP, and multi-electrode or single-unit recordings |
| genomics | DNA/RNA sequencing, variant data, single-cell omics, and multi-omics |
| biosamples | Physical biological material: brain tissue, biospecimens, CSF, cell lines |
| bioimaging | Light microscopy, electron microscopy, and biological imaging |
| behavior | Behavioural, cognitive, and psychophysiological measurements |
| clinical | Clinical trials, patient registries, and regulated research settings |
| health | Health data infrastructure, interoperability standards, and health data governance |
| computational | Computational models and simulations: model-description languages, morphology reconstruction formats, and the model and simulation repositories that produce or host them |
| reproducibility | Open code, workflow management, provenance tracking, and replication infrastructure |
Relationship predicates
Every connection between two nodes carries a predicate: a single word that says what kind of relationship it is. This page is the canonical list.
The predicates are not invented here. They form an application profile: a documented selection of terms drawn from established standard vocabularies, plus a few custom terms minted only where no standard covers the relationship. The standards reused are Dublin Core and schema.org for structural and descriptive relations, and FAIRsharing conventions for data-flow relations. Each is published as a typed, identified property, the way a W3C vocabulary such as DCAT defines dcat:distribution or SKOS defines skos:broader. Reusing their identifiers, rather than coining synonyms, is what lets the graph’s connections be matched to other systems later without translation. It is also the same discipline the graph documents elsewhere: a vocabulary supplies approved terms, and here the predicates are the vocabulary, while DCAT, SKOS, and Dublin Core are the published standards that define what each such term is.
Custom predicates are added only when the open-science governance relationships have no standard equivalent (for example governedBy, nationalNodeOf, correspondsWith). Where a standard term exists, it is preferred over a local one.
Predicates have a direction. The connection is always written on the more specific or dependent node, pointing toward the authority, standard, or parent it depends on. A repository writes endorsedBy: on its own page pointing at the body that endorsed it; the body never lists back. This keeps authority and parent nodes clean, since the reverse relationships appear automatically as backlinks.
The predicates fall into three groups by the kind of relationship they describe.
Technical: how data and tools relate
| Predicate | Meaning | Example |
|---|---|---|
accepts | The resource takes in data deposited in this format or standard | de.NBI accepts: NIfTI |
outputs | The resource delivers or exports data in this format | |
implements | The resource uses this standard internally, or runs on this platform’s software | Health Data Hub implements: OMOP CDM |
requires | A hard dependency: the resource cannot function without the target | OpenNeuro requires: BIDS |
extends | The resource builds on, derives from, or is a constrained profile of the target | NIDM extends: PROV-O |
The implements term also covers a repository deployed on a shared platform codebase, for example a data archive running on Dataverse or LORIS. The platform must itself be a node in the graph, and the edge is written on the repository pointing at the platform.
Structural: how entities belong to one another
| Predicate | Meaning | Example |
|---|---|---|
isPartOf | A structural sub-unit or component of the target | IFB isPartOf: ELIXIR |
memberOf | A formal member of a consortium or alliance that keeps its own independence | Rotterdam Study memberOf: ENIGMA Consortium |
fundedBy | The target is a primary funder | DANDI Archive fundedBy: NIH BRAIN Initiative |
producedBy | The target created or maintains this standard, resource, or tool | CIFTI producedBy: Human Connectome Project |
operatedBy | The target directly runs this digital resource | EDS AP-HP operatedBy: AP-HP |
nationalNodeOf | The designated national representative of a network | de.NBI nationalNodeOf: ELIXIR |
registeredIn | The entity’s trials, datasets, or identifiers are formally registered in the target, with a persistent identifier assigned | ADNI registeredIn: LONI IDA |
correspondsWith | Two peer standards share a jointly maintained mapping (written on both, pointing at each other) |
Depositing data in a repository counts as registeredIn when it results in a citable identifier such as a DOI or accession number, as it does for OpenNeuro, EGA, or DANDI Archive.
Governance: how rules and recognition relate
| Predicate | Meaning | Example |
|---|---|---|
governedBy | The target sets the rules, owns the specification, and controls its development | BIDS governedBy: BIDS Steering Group |
endorsedBy | The target formally recognises this resource without controlling it | BIDS endorsedBy: INCF |
recommendedBy | The target policy or body recommends this resource | Zenodo recommendedBy: EC Open Science Policy |
When nothing fits
| Predicate | Meaning | Example |
|---|---|---|
relatedTo | A direct relationship exists but no more specific predicate applies. Use sparingly, and add a short note in brackets to say what it is | relatedTo: GA4GH (DUO consent ontology) |
When no predicate fits and the relationship is weak, leave the link out rather than weaken the graph with an uninformative edge.
Adopting, aligning with, or advocating a standard or framework
A recurring case needs its own rule: an entity that adopts, aligns with, operates under, or advocates a standard, policy, or framework, without it being a hard technical dependency, a governance relationship, or a formal membership. There is no dedicated predicate for this, by design, since adding one would break the rule that every predicate maps to an established vocabulary. Route each case to the predicate that already fits:
- A hard technical use of the target (deployed on it, runs its software, cannot function without it) is
implementsorrequires. - A formal member of the body is
memberOf, and a structural sub-unit isisPartOf. - Mere compliance with a near-universal norm (adhering to FAIR, operating under a broad open-science policy that applies to a whole sector) is not an edge at all. It fails the consistency test in CONTRIBUTING (nearly every node would carry it), so it is recorded in prose, not the graph.
- A genuine, selective alignment that is none of the above is
relatedTowith a bracket note naming the alignment, for examplerelatedTo: [[FAIR Principles]] (builds its data model on FAIR). The note carries what a dedicated predicate would have, and reservesrelatedTofor relationships specific enough to be worth recording.
The test for the last two is the consistency test: if the alignment, applied consistently, would appear on nearly every node of the same kind, it is a universal property and belongs in prose. If it is selective, the annotated relatedTo records it.
ICM tags
Two additional tags mark the relationship of the Paris Brain Institute to nodes in the graph. They can be ignored by external visitors.
| Tag | Meaning |
|---|---|
| uses | The Paris Brain Institute operationally uses this standard, platform, or tool |
| participates | The Paris Brain Institute is an active member or participant in this body |