Making genomic analysis reproducible at research scale

Challenge

The challenge

Analysis pipelines had grown organically around individual researchers. Results depended on local environments, parameters were not captured consistently, and scaling sample volumes put growing pressure on shared infrastructure.

Context

The context

The scientific team had strong analysis expertise but limited software engineering capacity. The priority was to formalise what already worked, not to replace trusted tools.

Solution

Our approach

1. 01 Captured existing pipelines and their tool and parameter dependencies
2. 02 Reimplemented them as containerised, versioned workflows with recorded provenance
3. 03 Introduced orchestration and scheduling across on-premise and cloud compute
4. 04 Built a results interface for review and reporting rather than raw output directories
5. 05 Established data and metadata management so runs remain comparable over time

Implementation

How it was delivered

Pipelines were migrated one at a time and reconciled against existing results, so scientific confidence in the new process was established empirically before old workflows were retired.