The goal of Neuroshapes is the development of open, use case driven and shared validatable data models (schemas, vocabularies) to enable the FAIR principles (Findable, Accessible, Interoperable and Reusable) for basic, computational and clinical neuroscience (meta)data. The data models developed thus far entities for electrophysiology, neuron morphology, brain atlases, in vitro electrophysiology and computational modeling. Future developments could include brain imaging, transcriptomic and clinical form data, as determined by community interests.
Table of contents:
The main goal is to promote:
-
the use of standard semantic markups and linked data principles as ways to structure metadata and related data: the W3C RDF format is leveraged, specifically its developer friendly JSON-LD serialization. The adoption of linked data principles and JSON-LD will ease federated access and discoverability of distributed neuroscience (meta)data over the web.
-
the use of the W3C SHACL (Shape Constraint Language) recommendation as a rich metadata schema language which is formal and expressive; interoperable; machine interpretable; and domain agnostic. With SHACL, (meta)data quality can be enforced based on schemas and vocabularies (easily discoverable and searchable) rather than being fully encoded in procedural codes. SHACL also provides key interoperability capabities to ensure the evolution of standard data models and data longevity. It allows to incrementally build standard data models in term of semantics and sophistication.
-
the reuse of existing schemas and semantic markups (like schema.org) and existing ontologies and controlled vocabularies (including NIFSTD - NIF Standard Ontologies)
-
the use of W3C PROV-O recommendation as a format to record (meta)data provenance: a SHACL version of the W3C PROV-O is created.
Also, Neuroshapes aims at creating a community for an open and use case driven development of not only data models (schemas and vocabularies) and tools around them but also guidelines for FAIR neuroscience (meta)data.
A set of tutorials from the Blue Brain Nexus Forge project are available and use the schemas defined in Neuroshapes as data models to create and validate dataset as well as registering them in Blue Brain Nexus.
Try them in Binder
Neuroscience research generates vast amounts of data, but sharing and reusing this data has been challenging due to several issues:
- Lack of Standardization: Different research groups use various formats, making it difficult to integrate data from multiple sources.
- Limited Validation Mechanisms: Ensuring the accuracy and consistency of neuroscience data is a complex and manual process.
- Discoverability Issues: Researchers often struggle to find relevant datasets due to inconsistent metadata practices.
- Interoperability Challenges: Many neuroscience platforms and tools use incompatible data models, making it hard to exchange information.
Neuroshapes provides a structured, standard-based approach to neuroscience data management by:
- Using SHACL for defining and enforcing metadata rules.
- Adopting JSON-LD for semantic data structuring and discoverability.
- Leveraging Linked Data Principles to enable federated access to neuroscience datasets.
- Promoting reusability and interoperability, making data easily shareable across projects.
By addressing these challenges, Neuroshapes ensures neuroscience data is Findable, Accessible, Interoperable, and Reusable (FAIR).
The following projects have adopted Neuroshapes:
All schemas in this repository conform to the W3C SHACL recommendation and are serialized using JSON-LD.
Two different tests are executed in the unittest. The first test validates that schemas conform with the SHACL specifications.
The second tests consist of having valid and invalid data samples that are going to be tested against the modeled shapes. These examples are placed in the examples directory and follow the directory structure of the shape they should be tested against.
|-- examples
| |-- neurosciencegraph
| | |-- datashapes
| | `-- commons
| | `-- list
| | |-- schema.json
| | `-- examples
| | |-- datashapes.json
| | `-- valid
| | | `-- recipe_ingredients_list.json
| | `-- invalid
| | `-- recipe_missing_ingredients.json
| `-- prov
`-- ...
Tests require python > 3.6, and pytest. To run them follow next:
# create your virtual environment and activate it
python3 -m venv env
source env/bin/activate
# install requirements
pip install pytest pyshacl
# run tests
pytest
To test a set of shapes inside shapes directory, an optional argument can be used:
pytest --testdir=shapes/neurosciencegraph/datashapes/atlas
- Creation of an INCF/neuroshapes Special Interest Group
- INCF endorsement as a standard and best practice that support FAIR neuroscience data
- Extension of the current data model specifications
The license for all schemas and data is CC-BY-4.0.