01
Where it fits and where it does not
Use these four checks before committing implementation time.
- Use it when
- Reproducible bioinformatics and scientific workflows that must move across workstations, clusters, clouds, and compatible workflow engines.
- Limits
- Runner behavior outside the specified execution model can vary, and CWL alone does not freeze containers, reference data, credentials, resource policies, or scientific assumptions.
- Best for
- Omics and Bioinformatics and Computational modeling and AI / ML teams working across Plan → Harmonize → Exchange → Learn + reuse.
- Maturity
- EstablishedSuitable for production assessment. Pin the exact release and any implementation profile.
02
See it in the workflow
This view shows the input, the change the standard introduces, and the resulting output.
- InputWhat starts
Omics and Bioinformatics and Computational modeling and AI / ML source data, metadata, and local mappings
- CWLWhat changes
Use CWL as a pinned standard across Plan → Harmonize → Exchange → Learn + reuse
- OutputWhat becomes possible
A handoff the next system or team can validate against the same release
03
A concrete example
A genomics team validates a pinned CWL v1.2 workflow, container digests, reference checksums, inputs, outputs, and runner conformance before executing it through a WES endpoint.
Why it matters: Makes preprocessing and feature-generation workflows inspectable and portable, while dataset fitness, label validity, leakage controls, and run-level evidence remain separate gates.
04
What it fits with
CWL describes the executable workflow that GA4GH WES can run; BioCompute documents a reviewable analysis; RO-Crate can package workflow assets and run evidence; PROV-O can express provenance.
- Metadata vocabularyDPV
Both support Omics and AI / ML work and meet around Plan, Harmonize, Exchange, Learn + reuse. Compare their roles before treating them as interchangeable.
Explore relationship - StandardBioCompute
Both support Omics and Bioinformatics work and meet around Plan, Harmonize, Exchange, Learn + reuse. Compare their roles before treating them as interchangeable.
Explore relationship - Reference architectureAnnData ecosystem
Both support Omics and Bioinformatics and AI / ML work and meet around Harmonize, Exchange, Learn + reuse. Compare their roles before treating them as interchangeable.
Explore relationship - Ontology / data modelDUO
Both support Omics and AI / ML work and meet around Plan, Exchange, Learn + reuse. Compare their roles before treating them as interchangeable.
Explore relationship
05
Implementation starter
Start with one bounded handoff. Pin, test, and review it before scaling.
Define one handoff, its accountable owner, and the decision CWL must support.
Pin the exact version and companion artifacts: v1.2 · corrected specification text v1.2.1.
Map one representative input to the required standard artifacts.
Test the result against the canonical source and record every exception.
Preserve the source data, mappings, and review evidence before scaling.
06
Test the main limitation
Runner behavior outside the specified execution model can vary, and CWL alone does not freeze containers, reference data, credentials, resource policies, or scientific assumptions.
Run one representative end-to-end pilot and record exactly where CWL loses context, needs an extension, or depends on another standard.
Machine-readable output may still be unfit for analysis or ML.
Test the output for missing context, provenance, terminology alignment, time leakage, and the intended downstream decision. Makes preprocessing and feature-generation workflows inspectable and portable, while dataset fitness, label validity, leakage controls, and run-level evidence remain separate gates.
07
Official resources
Specifications, diagrams, examples, and guides from the organizations that maintain them.
CWL Workflow Description v1.2.1
Official publisher or steward guidance for this standard profile.
- Publisher
- Common Workflow Language working group · Software Freedom Conservancy