The Creator develops a workflow designed for maximum portability and registers the complete package as a persistent FDO artifact. This involves three main steps:

1. Develop a Snakemake pipeline with version-locked Conda environments for each rule
2. Package the workflow and its metadata as a Workflow RO-Crate archive
3. Register the archive in the MaRDI knowledge graph so it can be resolved and executed via its identifier

1. Develop the Snakemake Pipeline¶

The workflow must explicitly use the Snakemake Conda integration to ensure cross-platform reproducibility.

1. Write the Snakefile: Define the logic, rules, inputs, and outputs of your pipeline.

2. Define Isolated Environments: For every external tool or dependency, create a dedicated environment.yaml file listing version-locked dependencies (e.g., fastqc=0.11.9, python=3.9).

3. Link Environments in the Snakefile: Ensure every rule points to its environment file using the conda: directive.

   Example Snakefile snippet:

   rule fastqc_report:
       input:
           "data/{sample}.fastq"
       output:
           "results/{sample}_fastqc.html"
       conda:
           "envs/fastqc.yaml"
       shell:
           "fastqc {input} -o {output}"
   

4. Test for Portability: Test the workflow on Windows, Linux, and macOS using the --use-conda flag to confirm environments build correctly and execution is identical across systems.

2. Package and Register the FDO¶

The validated workflow files are packaged as a Workflow RO-Crate and registered in the MaRDI knowledge graph. The Workflow RO-Crate profile is an EOSC/ELIXIR standard for packaging executable workflows with enriched metadata, ensuring the package is interpretable by tools and communities beyond MaRDI without requiring registration in any external registry.

2.1 Assemble Workflow Artifacts¶

Collect the following into a single directory before crating:

• Snakefile — the main workflow definition
• envs/ — all environment.yaml files
• Any configuration files
• README.md — description of the workflow, inputs, outputs, and parameters
• CITATION.cff — citation metadata for proper attribution
• test/ — test input data and expected outputs (see section 3)

2.2 Generate the Workflow RO-Crate¶

From inside the workflow directory, use the rocrate CLI (included in requirements.txt):

rocrate init
rocrate add workflow -l snakemake Snakefile
rocrate add file README.md
rocrate add file CITATION.cff
rocrate write-zip workflow_A01.crate.zip

Richer metadata — such as creator ORCID, Snakemake version, or parameter descriptions — can be added by editing the generated ro-crate-metadata.json directly before packaging.

2.3 Register in the MaRDI Knowledge Graph¶

A workflow FDO has two parts: the RO-Crate archive is stored in MaRDI storage, while the FDO metadata — the item that makes the workflow findable and resolvable — lives in the MaRDI knowledge graph, which serves as a registry for datasets and research artifacts from the broader community.

Registration involves two steps:

1. Create a knowledge graph item for the workflow (if one does not already exist), recording its metadata (creator, description, parameters, Snakemake version). This gives the workflow a stable QID.
2. Upload the RO-Crate to MaRDI storage under the workflow's QID with component ID rocrate.

Once uploaded, the MaRDI FDO server automatically detects the file and lists it in the FDO manifest under fdo:hasComponent. The MaRDI DOIP server can then resolve and serve the workflow by its QID without any further steps.

Next: Workflow Testing | Executing a Workflow