hpci improves reproducibility of computational biology research by integrating Continuous Integration with High Performance Computing clusters

Abstract

Abstract Motivation The reproducibility crisis in bioinformatics and computational biology can be mitigated by using Continuous Integration (CI), a proven software engineering practice, to automate the testing of analysis pipelines and models. However, the adoption of CI in bioinformatics is limited as existing solutions are typically tightly linked to specific CI platforms, specific HPC job schedulers, or require actions from HPC administrators. Results We developed hpci, a platform-agnostic, open-source utility that integrates CI platforms (e.g., GitHub Actions, GitLab CI) with common HPC job schedulers (Slurm, PBS). hpci operates as a standard unprivileged user via SSH, and requires no cluster-side installation. It terminates on the CI runner using the exact same exit code generated by the HPC job, providing feedback to researchers and preventing downstream CI steps if there are errors. We demonstrate hpci’s utility across five diverse computational biology workflows on HPC, highlighting the value of automated testing and continuous analysis for verifying reproducibility in bioinformatics workflows. Availability hpci is implemented in Haskell and is available under the GPL-3.0 licence. Source code and binaries can be downloaded from https://github.com/Garvan-Data-Science-Platform/hpci. Source code is archived in Software Heritage via Zenodo (https://zenodo.org/records/19587650).