Skip to content


This glossary compiles some key terms used in the Neurobagel documentation and defines them in the context of the Neurobagel ecosystem.

Data dictionary

A JSON file that describes the information contained in columns from a tabular data file, along with the meaning and properties (format of numerical data, unique “levels” of categorical data, etc.) of values in each column. In the context of Neurobagel, the meanings of columns and column values are encoded using terms from standardised vocabularies.

Data model

Used interchangeably with: data schema

A structure that has been designed with the purpose to represent a specific kind of information. A data model is made up of generic types or classes that are relevant to the data model designers (for Neurobagel, examples include "Research Participant" and "Neuroimaging Dataset"), the properties these types can have (e.g., "Age in years", "Dataset name"), and the relationships that can exist between them (e.g., "is part of"). The goal of a data model is to give information a structure so that we can write programs that can consume the information.

The Neurobagel data model is designed to represent the kind of information that is important to support the most relevant cohort definition queries, and thus models types, properties, and relationships that are important for this purpose. It is not a static thing, and we constantly add new things to the data model as we support new use cases that rely on this information.

Controlled term

A unique identifier or code for a concept that is described in a controlled vocabulary.

A controlled term has a

  • a clear definition
  • a unique and persistent identifier
  • from a specific curated list of terms like a vocabulary, taxonomy or ontology

An example is the controlled term for "Parkinson's disease" from the ICD-11 taxonomy with the unique code 8A00.0.

Controlled vocabulary

Used interchangeably with: taxonomy, and ontology

A controlled vocabulary is a collection of controlled terms that are often all about one specific topic. The main benefit of a controlled vocabulary is that it provides unambiguous terms with clear definitions that people have agreed to use to describe their information - removing the need to align variable names and value formats between datasets and enabling interoperability.

For example, most websites use the vocabulary to describe things like products to purchase, events to book, recipes to cook etc. in a consistent way that can be understood by the search spiders of big search engines.

Reusing controlled vocabularies

Creating a controlled vocabulary is a laborious task that involves deep subject matter expertise, often from many experts, and needs to be maintained to remain relevant. You should therefore almost always reuse an existing vocabulary rather than creating your own.

A taxonomy is a more specific form of a controlled vocabulary that organizes terms into hierarchical relationships. For example, a "Recipe" in is a subtype of a "HowTo" which itself is a subtype of a "CreativeWork". This hierarchy let's you do things like search for "CreativeWork" and also find "Recipe", even if you have never made this link directly.

An ontology is an even more specific form of a taxonomy where terms can have very complex relationships with each other that include logical constraints. In an ontology, you could for example express that for someone to be a "sister" to someone else, both the subject and the object of the relationship have to be "human", only the subject of the relation has to be "female", and both have to have at least one parent in common. These complex expressions are very labour intensive to create but can provide also very rich ways of validating and even inferring information.

Graph database

Used interchangeably with: knowledge graph store, graph store, graph

A type of database, in the same way that a relational databases is a type of database. The main distinguishing feature of graph databases is that they represent entities as nodes in a graph, and relationships between entities as edges between these nodes. This data model makes it easy to easily add new information by drawing a new edge between two nodes.


A single Neurobagel graph database can contain harmonised information about multiple datasets and their respective subjects. Each subject is represented by a node, and their harmonised phenotypic and imaging data characteristics are described using controlled terms connected to the subject node via a series of edges that individually encode the type of attribute described by the controlled term.

Neurobagel uses the RDF graph data model, see also


In the context of Neurobagel, annotation refers to the process of describing tabular demographic, cognitive, and/or clinical (phenotypic) data for a dataset with terms from controlled vocabularies to create machine understandable data dictionaries for the data. You can learn more about this process in our documentation.

Aggregated results

If the owner of a Neurobagel node decides that query responses should not include information at the level of individual participants, they can configure their node to only return aggregated results. In this mode, the node will aggregate all participants that match a query at the dataset level and only respond with counts of matching participants.

Data owner

A person or an institute who is responsible in the data governance sense for one or many datasets. In the context of Neurobagel, one data owner can have one or more Neurobagel nodes, but every Neurobagel node can only have one data owner who is responsible for all of the data stored inside the node.

Federation API

Used interchangeably with: f-API

A standalone service that allows query users to send a single query and have it automatically sent to many Neurobagel node APIs (n-API) without having to know where these node APIs are located. The f-API takes care of keeping an up to date list of available n-APIs, federating queries, retrieving and combining results, and returning them to the user.

Designed to very closely resemble the behaviour and the endpoints of a n-API so that services can be built that are able to work either directly with a single n-API or with an f-API.

Node API

Used interchangeably with: n-API

A Neurobagel "node" is a locally deployed service that holds information about data for one data owner who controls and manages the node. A node has two core components:

  • a graph backend to store the harmonised data for querying
  • a RESTful node API that exposes query endpoints for users or programs to send queries and retrieve results

One important purpose of the n-API is to act as a barrier between the user and the graph backend so that the user cannot execute arbitrary queries on the graph, and the data owner can control how detailed the query responses should be.

Tabular data

Used interchangeably with: phenotypic data

Tabular text files (e.g., .tsv or .csv) that contain information about participants such as their demographic information or data from cognitive or clinical assessments they have completed. We often refer to this information as phenotypic data because they describe observable characteristics of the participant.