1. Validate the service intake
The first process checks that the sample sheet contains exactly eight unique samples, four complete untreated/treated pairs, the correct accession and the declared processed-count input type.
Bioinformatics workflow
Reproducible RNA-seq Service Workflow
I built a compact Nextflow workflow around a paired differential-expression question in the public GSE52778 airway dataset. The emphasis is the complete service path: a precise intake, an explicit statistical design, checked outputs and a handover another person can inspect.
Quick scan
A paired analysis with a declared contract
The workflow begins with the processed counts distributed through Bioconductor and keeps the biological question, pairing and contrast visible from intake to handover.
- Project type
- Self-directed bioinformatics service-workflow exercise
- Public dataset
- GSE52778 through the Bioconductor airway package
- Analysis unit
- Eight processed-count samples forming four untreated/treated cell-line pairs
- Model
- ~ cell + dex, reporting treated versus untreated
- Workflow
- Intake validation, paired DESeq2 analysis, result checks and handover bundle
- Reproducibility
- Nextflow DSL2, pinned Bioconductor container, provenance hashes and CI-tested validators
Service path
Four stages from intake to handover
Nextflow makes the dependency between stages explicit while each validator keeps a separate, readable contract.
2. Run the paired analysis
DESeq2 fits the declared ~ cell + dex design, applies a fixed low-count rule and reports the dexamethasone-treated versus untreated contrast while controlling for the four cell-line pairs.
3. Check the result contract
A separate validator checks result columns, adjusted-p-value filtering, model metadata and the expected positive CRISPLD2 direction so a reversed contrast or incomplete handoff is caught explicitly.
4. Build the handover
The final process assembles result tables, QC plots, metadata, session information, interpretation notes and a provenance manifest containing software versions and SHA-256 hashes.
Why the pairing matters
The samples come from four cell lines measured once untreated and once after dexamethasone treatment. Using ~ cell + dex keeps those baseline differences in the model instead of treating eight related measurements as independent samples.
What the workflow starts from
This project begins with processed gene counts from the airway package. Raw FASTQ quality control, alignment and quantification would be a separate upstream service, so they are not implied by the current workflow.
Reproducibility
The handover is part of the analysis
The project records the dataset accession, Git commit, container references, package versions and file hashes alongside the result tables and plots.
What this project adds
It extends my biomedical data work into a service-style bioinformatics workflow where the analysis is only one part of the job. Intake quality, contrast direction, result completeness and a usable handover all need to survive review as well.
Page updated