Skip to content

nf-osi/nf-metadata-dictionary

Repository files navigation

GitHub release (latest by date) GitHub Release Date GitHub

Maintenance and Contribution

This data model, maintained by the NF-Open Science Initative, provides standard concepts and structure to help describe and manage data and other resources in the NF-OSI community.

One can reference the data model to understand things such as:

  • What type of entities there are (e.g. RNA-seq files vs image files, files vs datasets)
  • What metadata properties exist for these different types, to help understand and use a data resource
  • What are the preferred/standard labels for something as defined by community input and our data managers (e.g. prefer "NF2-related schwannomatosis" vs the now-deprecated "Neurofibromatosis type 2")

Terms in the metadata dictionary are used in the manifests within the Data Curator App and on the NF Data Portal. We welcome contributions from community members, whether you are a professional data modeler, clinician, or student in the NF community.

Data Model Framework

The data model is maintained as a subset of the YAML-based Linked Data Modeling Language (LinkML) that is compatible with our internal tool schematic. This subset of LinkML should be easy to get started with.

The 10-minute Intro

The data model primarily models different types of biological data and patients/samples and uses templates that collects information for these entities. (It may helpful to also read about Entity-Relation Model.)

To do this the building blocks are:

LinkML term Schematic Note Other Note
Class Usually corresponds to "Component"; schematic uses definition to make a "manifest" Also known as "template"
Slot Schematic calls these "attributes"; the range can be typical types (string, int, etc.) or enumerations (below) Also known as "property"
Enum (enumeration) Schematic calls these "valid values" Also known as "controlled values"

Classes depend on slots being defined, and some slots depend on enumerated values being defined. If a class uses a slot that is not defined, the model will error when trying to build; and same with a slot that uses an enumeration. Slots can be reused across classes; and same with enumerations for slot. For example, unit can be reused across any class entities that needs to capture unit information.

Classes

Classes have slots (properties). All classes are grouped under modules/Template. Classes can be built upon, so subclasses inherit properties from a parent class.

Example: Base patient class

Here is an small-ish base class definition for a patient:

  PatientTemplate:
    is_a: Template
    description: >
      Template for collecting *minimal* individual-level patient data. 
    slots:
    - individualID
    - sex
    - age
    - ageUnit
    - species # should be constrained to human
    - diagnosis
    - nf1Genotype
    - nf2Genotype
    annotations:
      requiresComponent: ''
      required: false

Slots

Slots have a range, which can be a basic data type such as "string" and "integer", or a set of controlled values (see Enum). If the range of a slot is not explicit defined, it defaults to "string" (basically free text). All slots are in the file modules/props.yaml.

Example slots
slots:
  #...more above
  specimenType:
    description: Association with some tissue (a mereologically maximal collection of cells that together perform physiological function).
    range: SpecimenTypeEnum # Take a look at SpecimenTypeEnum below
    required: false
  title:
    description: Title of a resource.
    meaning: https://www.w3.org/TR/vocab-dcat-3/#Property:resource_title
    required: true
  # ...more below

Here, required defines globally whether this slot must be filled out. For example, specimenID is required, and is included for assay data class definitions that involve a specimen. For that reason, it is not included at all in individual-level assay data (e.g. behavioral/psychological data). Note: In situations where "the data meets the template", issues with a required slot might indicate that the template being used doesn't actually "fit" the type of data/entity, and an additional one needs to be defined.

Enum

An enumeration is a set of controlled values. Enums are most of the files modules, everything except for what's in Templates and props.yaml.

Example: SpecimenType enumeration
enums:
  SpecimenTypeEnum:
    permissible_values:
      cerebral cortex:
        description: The outer layer of the cerebrum composed of neurons and unmyelinated nerve fibers. It is responsible for memory, attention, consciousness and other higher levels of mental function.
        meaning: http://purl.obolibrary.org/obo/NCIT_C12443
      bone marrow:
        description: The soft, fatty, vascular tissue that fills most bone cavities and is the source of red blood cells and many white blood cells.
        meaning: http://purl.obolibrary.org/obo/BTO_0000141
     #...more below
Example: Boolean enumeration (commonly reused)
enums:
  BooleanEnum: 
    description: 'Boolean values as Yes/No enums'
    permissible_values:
      'Yes':
        description: 'True'
      'No':
        description: 'False'

General meta to describe classes, slots, and enums

Aside from meta specific to each type (class, slot, or enum) above, terms have common meta, where the most prominent are summarized here:

  • description: Description to help understand the term.
  • meaning: This is a highly recommended and should be an ontology URI that the term maps to.
  • source: This can be used to supplement meaning, but it's most often used when an ontology URI does not exist. It provides a reference such as a publication. For example, a very novel assay might not have a real ontology concept yet but will likely be described in a paper.
  • notes: Internal editor notes.

Steps to add an attribute to the Metadata Dictionary:

  1. Create a new branch in the NF-metadata-dictionary repository. (example: patch/add-attribute)

  2. Find the yaml file in the new branch where the attribute belongs. The components of the data model are organized in the folder labeled modules.

  3. Create a pull request (PR) to merge the branch to "main". Add either @allaway, @anngvu, or @cconrad8 as a reviewer. Creating the PR will:
    i. Build the NF.jsonld from the module source files. This is the format needed by schematic. Therefore, you do not need to edit the NF.jsonld directly, because it will be done automatically by our service bot.

    ii. If build succeeds, also run some tests to make sure all looks good/generate previews. After some minutes, a test report will appear in the PR that hopefully looks like this:

    image
  4. Make any necessary changes and then merge the new branch that was created to the main branch.

  5. Draft a versioned release here.

  6. Name the release with the convention MAJOR.MINOR.PATCH. These releases start at v1.0.0. Versioning is roughly following semantic versioning concepts where:

    • MAJOR: In-use parent attributes are deleted from dictionary or modified, or in-use child attributes are modified in a non-backwards compatible way (e.g. Neurofibromatosis 1 changed to Neurofibromatosis type 1).
    • MINOR: Concepts/parent attributes are added.
    • PATCH: Child attributes are added, or unused child/parent attributes are deleted/modified, or definitions/comments are added/modified, or validation rules are modified in a backwards compatible way.
  7. Navigate to data curator app config file to update the version. First, update the staging branch to test on the staging app. Then, checkout the main branch on your fork, update dcc_config.csv for NF so it matches the staging set up, then create a PR to main on the sage-bionetworks repo. Once merged, the changes should be visible on the data curator app.

  8. 🎉 Congrats! The term is now added to the metadata dictionary.

Further Information

Building Locally

To build locally, install the schematic package.

Help

For questions or to get help contributing a term, please create an issue.

License

The "collection" of metadata terms in this repository is made available under a CC0 license. The individual terms and their definitions may be subject to other (permissive) licenses, please see the source for each metadata term for details.