Computational workflows are simultaneously software, something executable, and data, a specification of a process. Curating and managing them are increasingly part of your task as data steward. Dependencies become deprecated and containers become unavailable, making it more difficult to reproduce a workflow with time. Even when it still runs, you need detailed provenance to know whether an output came from the same code, the same version, and the same parameters.
In 2025, the FAIR Computational Workflows Working Group (part of the Workflows Community Initiative) published guidelines adapting FAIR to computational workflows (see resources below). This page translates those guidelines into practical activities you can perform with existing tools. The activities follow a workflow’s life cycle, but you can apply each on its own. Tool tips use Nextflow, but the principles also apply also CWL, Snakemake, Galaxy, and other workflow engines.
Make sure each workflow is registered in a searchable registry
When you curate a workflow, check that it is deposited in WorkflowHub or Dockstore (see resources), and help the authors register it if it is not.
Don’t rely on a GitHub repository alone
A GitHub repository alone is not indexed in a searchable, citable way, registration is what makes the workflow findable to others.
Use a DOI issued by the registry
Both registries mint persistent identifiers (DOIs) for registered workflows, store metadata, and track versions.
Set up an institutional space in WorkflowHub
If your institution or project has many workflows, consider setting up an institutional or project space in WorkflowHub so they are grouped and easy to find.
Check that releases are versioned and cited with a version-specific DOI
Confirm that each release is tagged in version control and registered (see previous section), so that every version has its own DOI. Different versions of a workflow can produce different results, so collaborators and reviewers need to cite the exact version used.
Issue a DOI for each release
In WorkflowHub, a DOI can be generated per tagged release; connecting a GitHub repository to Zenodo achieves the same for simpler cases.
Review and complete the workflow metadata
Check that the workflow is described with machine-readable metadata, and fill gaps with the authors.
Use metadata standards
The Bioschemas ComputationalWorkflow profile or CodeMeta captures title, authors with ORCIDs, workflow specification language, licence, and dependencies.
Fill in the Nextflow manifest block
For Nextflow, this metadata is recorded in the manifest block in nextflow.config.
Check the licence of the workflow and its components
Make sure the workflow as a whole has a licence, then verify it is compatible with the licences of all its components, tools, sub-workflows, and containers each carry their own.
Check for incompatible licences
Incompatible licences can restrict how the combined output is shared. This is easy to overlook when assembling modular workflows.
Record runs, not just the workflow specification
Check whether the workflow engine was configured to record which input files and parameters were used, which tool versions ran, and what outputs were produced.
Support reproducibility
Provenance is the evidence you need to reproduce a specific execution, not the workflow specification alone, but the record of a run on specific inputs.
Enable nf-prov or Nextflow lineage in your workflow
For Nextflow, enable the nf-prov plugin, which records provenance as a Workflow Run RO-Crate or a BioCompute Object, or the data lineage option in the Nextflow configuration. nf-prov and Nextflow lineage capture are not enabled by default, they are opt-in configurations. Build them into your workflow template so new pipelines inherit them from the start.
Archive the workflow as an RO-Crate
For long-term preservation, make sure the workflow, its metadata, and (when feasible) an example run are packaged into an RO-Crate, a packaging standard that keeps the workflow readable and citable even when the execution platform is no longer available (see resources).
Package Nextflow workflows into RO-Crates
For Nextflow, nf-core pipelines rocrate packages the workflow with Bioschemas-compatible metadata that can then be uploaded to WorkflowHub for registration.
External resources
| Category | Resource | Description |
|---|---|---|
| External | Applying the FAIR Principles to Computational Workflows | FAIR guidelines for workflows from the WCI FAIR Computational Workflows Working Group (Scientific Data, 2025). |
| External | WorkflowHub | FAIR workflow registry supporting DOI minting, RO-Crate packaging, Bioschemas metadata, and versioning. |
| External | Dockstore | Open workflow registry widely used in genomics; supports CWL, WDL, Nextflow, and Galaxy. |
| External | RO-Crate | Lightweight packaging standard for bundling a workflow with its metadata and provenance records. |
| External | RO-Crate Workflow Run profile | Profile for packaging the record of a specific workflow execution alongside the workflow. |
| External | nf-prov (Nextflow) | Nextflow plugin that captures run-level provenance and emits a Workflow Run RO-Crate or BioCompute Object. |
| External | Getting started with data lineage | Data lineage in Nextflow provides tracking of workflow runs, task executions, and output files. |
| External | Bioschemas ComputationalWorkflow profile | Structured metadata schema for describing computational workflows in a machine-readable way. |