Ensuring reproducibility

A key goal of the OpenScPCA project is to ensure that results remain reproducible over time. This section describes steps that we take to achieve this goal and ensure a stable codebase, illustrated by the figure below:

• We require each analysis module to have its own software environment with pinned dependencies
  • As analysis modules mature, these software dependencies will be added to a module-specific Docker image
• We use the GitHub Actions continuous integration framework to automate certain processes that ensure reproducibility, including:
  • Testing that analysis modules run to completion without errors
  • Building module-specific Docker images and pushing them to a public registry for broad reuse
• We maintain a separate repository, OpenScPCA-nf, where completed analysis modules are run in a reproducible Nextflow workflow to generate workflow results associated with OpenScPCA data releases