A Python tool to automatically assemble models and tools into workflows to process SPARC datasets in accordance with FAIR principles.
- About
- Introduction
- The problem
- Vision and benefits
- Our solution - sparc-assemble
- Designed to enable FAIRness
- Future developments
- Setting up sparc-assemble
- Using sparc-assemble
- Reporting issues
- Contributing
- Cite us
- License
- Team
- Acknowledgements
This is the repository of Team sparc-assemble (Team #2) of the 2024 SPARC Codeathon. Click here to find out more about the SPARC Codeathon 2024. Check out the Team Section of this page to find out more about our team members.
Please see the Acknowledgements section of this readme for a list of tools that have been used in this project.
The NIH Common Fund program on Stimulating Peripheral Activity to Relieve Conditions (SPARC) focuses on understanding peripheral nerves (nerves that connect the brain and spinal cord to the rest of the body), how their electrical signals control internal organ function, and how therapeutic devices could be developed to modulate electrical activity in nerves to improve organ function. This may provide a potentially powerful way to treat a diverse set of common conditions and diseases such hypertension, heart failure, gastrointestinal disorders, and more. 60 research groups spanning 90 institutions and companies contribute to SPARC and work across over 15 organs and systems in 8 species.
The SPARC Portal provides a single user-facing online interface to resources that can be shared, cited, visualized, computed, and used for virtual experimentation. A key offering of the portal is the collection of well-curated datasets in a standardised format, including anatomical and computational models that are being generated both SPARC-funded researchers and the international scientific community. These datasets can be found under the "Find Data" section of the SPARC Portal. Information regarding how to navigate a SPARC dataset and how a dataset is formatted can be found on the SPARC Portal.
The scientific community is developing new tools and models to process and understand measurements in these datasets to generate new results, outcomes, and knowledge. These tools and models are often applied in a series of steps to create workflows whose outputs provide quantities of interest. For example, a user may be creating a workflow that inputs medical images of the brain (measurement) and segments brain tissue from these images (tool) that can output brain volume (quantity of interest). Another user maybe interest in a workflow that inputs electrode measurements from the surface of the heart to personalise a computational model of cardiac electrophysiology to quantify activation patterns. These workflows may require multiple intermediate tools and models to generate the final output of interest.
During the past decades, tens of thousands of models and tools have been developed globally from different initiatives and stored in multiple online repositories, for example:
- Models
- Physiome Model Repository (>6,000 models)
- Biomodels (>3,000 models)
- Tools
- biotools (>30,000 tools)
- OpenEBench (>40,000 tools)
- BioContainers (>100,000 tools)
- Galaxy (>9,000 tools)
- WorkflowHub (>700 tools)
Despite users having a general idea of the quantities of interest in their investigations (e.g. outputs of specific models or tools they are developing), they typically need to assemble workflows manually, often guessing what inputs and intermediate measurements, models, and tools may be needed.
There is currently no option for users to:
- easily find and access existing models and tools:
- developed by the SPARC community for processing SPARC data
- developed externally that could be used for processing SPARC data
- easily assemble new workflows that reuse existing models and tools to evaluate quantities of interest
- easily identify which measurements (e.g. SPARC datasets) already contain the necessary inputs for these new workflows
- easily identify which measurements, tools, and/or models may be missing to support efficient advancement of research efforts
Our vision is to:
- providing a knowledgebase that:
- describes existing measurements, models, and tools developed by the SPARC and wider communities
- provides a map of all the inputs and outputs of the available models and tools
- leverage the knowledgebase to automatically assemble workflow descriptions to evaluate quantities of interest
- run the assembled workflow(s) or help identify missing components that are needed to run the assembled workflow
- store assembled workfow descriptions in a FAIR manner such that they can potentially be e.g. contributed to the SPARC Portal
Providing these capabilites would:
- significantly improve resource search functionality, especially if it is integrated with shared infrastructure such as the SPARC portal. This would allow users to find existing tools and models that may already enable evalaution of some or all of the quantities that they are interested.
- maximise finability, reusabilty, and therefore, the impact of existing SPARC resoures (providing an pathway for other communities that are building tools and models to make use of SPARC data)
- support reuse of assembled workflows by the community for generating scientific advances.
- help the community identify gaps in our knowledge and capabilites to support and help prioritise future research developments
To address this problem and support our vision, we have developed a Python module called SPARC Assemble (sparc-assemble) that can be used to find, access, and automatically assemble models and tools into workflows to process SPARC datasets in accordance with FAIR principles.
The following features are currently supported:
- Extracting and annotating existing tools and models from SPARC datasets to help standardise and harmonise their input and output descriptions
- Accessing existing tools and models from external repositories such as WorkflowHub and Biomodels
- Storing information about the model and tools in a local knowledge graph defined by the standardised EDAM ontology (which has been deveoped for bioscientific data analysis and data management)
- Listing all available models and tools
- Performing queries to automatically assemble workflows:
- List all possible workflows (model and tool combinations) that would enable evaluation of a quantity of interest (ie an output of a model or tool)
- Identify which SPARC datasets contain the required inputs to the workflow
- Idenitfy which measurement, model, and tool inputs are missing
- Stores assembled workflows using the SPARC Dataset Structure to ensure they are FAIR
- Provides an easy-to-use python-based application programming interface (API) to provide the above functionality
- Provides a natural language inteface to make it easy for users to specify their quantity of interest
- Provides a series of tutorials to demonstrate the functionality of sparc-assemble
- Reuses existing SPARC resources and tools including sparc-me, sparc-flow, and the sparc-python-client.
We have assessed the FAIRness of our sparc-assemble against the FAIR Principles established for research software. The details of this assemsement is available in the codeathon google drive for team 2.
Additionally, sparc-assemble has adopted exsiting dataset, knoweledge graph and workflow standards including:
- The Common Workflow Language (CWL) - is an open standard and specification used in the field of bioinformatics and scientific computing to describe and execute tools and workflows. CWL provides a way to define and share complex computational tasks and data processing pipelines in a portable and platform-independent manner. It uses a JSON-based format to describe input data, processing steps, and output data, allowing researchers to collaborate and share reproducible analyses across different computing environments. CWL aims to enhance the ease of defining, sharing, and executing computational workflows, particularly in the context of data-intensive scientific research.
- EDAM ontology - EDAM is a comprehensive ontology of well-established, familiar concepts that are prevalent within bioscientific data analysis and data management (including computational biology, bioinformatics, and bioimage informatics). EDAM includes topics, operations, types of data and data identifiers, and data formats, relevant in data analysis and data management in life sciences.
- SPARC Dataset Structure
- Standardise the description of intputs and outputs of these models and tools
- Integrate our knowledge graph with other knowledgebases including the SPARC Anatomical Connectivity Maps and SPARC Functional Connectivity maps. This will enable workflows to be automatically assembled not only based on input/output relationships, but also based on anatomical and physiological connectivity.
- Expand tool descriptions that can be accessed e.g. Workflow Description Language, Nextflow, Snakemake etc
- Link with Large Language Models to support more complex queries, for example to help visualise quantities of interest.
- Show how the assembled workflows can be run with sparc-flow directly from the commandline or through existing cloud computing platforms from Dockstore.org (currently supports running on AnVIL, Cavatica, CGC, DNAnexus, Galaxy, Nextflow Tower, and [Terra].
- Provide API documentation.
- Git
- Python. Tested on:
- 3.10
- Operating system. Tested on:
- Ubuntu 24.04
Here is the link to our project on PyPI
pip install sparc-assemble
Clone the sparc-assemble repository from github, e.g.:
git clone https://github.com/SPARC-FAIR-Codeathon/2024-team-2.git
-
Setting up a virtual environment (optional but recommended). In this step, we will create a virtual environment in a new folder named venv, and activate the virtual environment.
- Linux
python3 -m venv venv source venv/bin/activate- Windows
python3 -m venv venv venv\Scripts\activate -
Installing dependencies via pip
pip install -r requirements.txt
If you find sparc-assemble useful, please add a GitHub Star to support developments!
Guided Jupyter Notebook tutorials have been developed describing how to use sparc-assemble in different scenarios:
| Tutorial | Description |
|---|---|
| 1 | Annotation of computational models and tools from existing SPARC datasets. |
| 2 | Building a knowledge graph for automated workflow assembly. |
| 3 | Adding tools created from existing SPARC datasets to the knowledge graph. |
| 4 | Adding non-SPARC tools and models to the knowledge graph. This includes adding an existing model from the Biomodels repository, and a tool from the WorkflowHub repository. |
| 5 | Assembling a workflow automatically using sparc-assemble and natural language processing |
| 6 | Running assembled workflows |
To report an issue or suggest a new feature, please use the issues page. Issue templates are provided to allow users to report bugs, and documentation or feature requests. Please check existing issues before submitting a new one.
Fork this repository and submit a pull request to contribute. Before doing so, please read our Code of Conduct and Contributing Guidelines. Pull request templates are provided to help guide developers in describing their contribution, mentioning the issues related to the pull request and describing their testing environment.
/sparc_assemble/- Parent directory of sparc-assemble python module./sparc_assemble/core/- Core classes of sparc-assemble./resources/- Resources that are used in tutorials (e.g. SDS datasets containing workflow and tool descriptions)./tutorials/- Parent directory of tutorials for using sparc-assemble./development_examples/- Parent directory of examples that were created during the development of sparc-assemble./docs/images/- Images used in sparc-assemble tutorials.
If you use sparc-assemble to make new discoveries or use the source code, please cite us as follows: Please note that the Zenodo link is a placeholder and can only be added once this repository is made public (after the codathon is completed).
Mathilde Verlyck, Jiali Xu, Max Dang Vu, Thiranja Prasad Babarenda Gamage, Chinchien Lin (2024). sparc-assemble: v1.0.0 - A Python tool to find and automatically assemble tools and workflows to process SPARC datasets in accordance with FAIR principles. Zenodo. https://doi.org/XXXX/zenodo.XXXX.
sparc-assemble is fully open source and distributed under the very permissive Apache License 2.0. See LICENSE for more information.
- Mathilde Verlyck (Developer, Writer - Documentation)
- Jiali Xu (Developer, Writer - Documentation)
- Max Dang Vu (Developer, Writer - Documentation)
- Thiranja Prasad Babarenda Gamage (Writer - Documentation)
- Chinchien Lin (Lead, SysAdmin)
We would like to thank the organizers of the 2024 SPARC Codeathon for their guidance and support during this Codeathon.
This Codeathon entry leveraged functionality of an existing software tool that provided the basic functionality of adding a tool to a knowledge graph (which used the EDAM ontology) and automated workflow assembly. Please note that this repository is currently private. Please provide the github IDs of the judges and access will be provided.
No work was done on this codeathon project prior to the start of the codeathon. All additional functionality described in the "our solution" section above was implemented during the codeathon including refactoring and incorporating the core functionality into a library and API within sparc-assemble, extending it to cater for models, natural language processing, all tutorial content etc, as evidenced by our commit history.