# Case study: From laptop scripts to a production pipeline
URL: https://omicsquest.com/case-studies/startup-pipeline-rebuild/

When the only computational scientist left, nobody could reproduce last quarter's results. We rebuilt the analysis as a tested, versioned pipeline the team owns.

Problem: Analysis lived in undocumented R scripts on one laptop, with hard-coded paths and package versions that no longer installed. After that scientist left, the team could not reproduce figures shown to its board, and new data piled up.

Findings:
- No version control, no environment pinning, and no record of which parameters produced which results.
- Inconsistent QC thresholds between runs made comparisons across experiments meaningless.

Rescue:
- Built a modular Nextflow pipeline on nf-core components, with pinned containers, for both scRNA-seq and bulk RNA-seq.
- Deployed on AWS Batch with cost controls, and storage lifecycle rules for raw data.
- Added automated QC reports, data versioning and a results catalog so every figure traces back to inputs and code.
- Wrote tests on small reference datasets and continuous integration so updates cannot silently change results.
- Moved to monthly flat-rate maintenance: version upgrades, new modules, monitoring, and a duo team on call for analysis.

Result: Each new sequencing run now produces a QC report and standard analysis within two days, without manual work. The team reproduced its historical results and passed due diligence for its next financing round.

---
Source: https://omicsquest.com/case-studies/startup-pipeline-rebuild/ | OmicsQuest, contact@omicsquest.com