Associate Director, Clinical Pharmacology Modeling & Simulation, Infectious Diseases & Vaccines GSK Weatogue, Connecticut, United States
Disclosure(s):
Janelle Lennie: No financial relationships to disclose
Objectives: Performing standardized, traceable, and documentable quality control (QC) on scripts is essential in data analysis and regulatory submissions. Current processes often rely on collaborative-editing software such as Microsoft Office applications or Atlassian Confluence. These workflows suffice in documenting QC completion but lack specific functionalities to facilitate QC workflows effectively. The open-source R [1] package ghqc was developed to leverage GitHub's version control and UI to accommodate QC's dynamic nature, organizing discussions and orienting both authors and QCers throughout the QC process. This work aims to demonstrate the capabilities of ghqc.
Methods: The proposed QC workflow centralizes meta-data, tracking, discussion, and decision-making on GitHub, while R Shiny apps are utilized for automating QC operations and importing meta-data to GitHub.
Within a GitHub repository, the ghqc "assign" app is utilized by the Author to automate the initiation of a QC review for one or more files at a named git commit and branch. The files to be reviewed are organized into GitHub Milestones [2], in which each file corresponds to a GitHub Issue [3]. The “assign” app automatically creates the body of an Issue which provides file context, including authors, QCers, commits, branches, editors, and links to the file’s contents. The file-specific checklist to guide QC is also included in the Issue body for dynamic completion during QC.
Results: Ghqc (v0.4.13) [4] was tested in real QC workflows. The interface proved intuitive, enabling Authors and QCers to manage QC activities seamlessly with minimal training done via brief demonstrations. Teams successfully assigned, discussed, tracked and resolved QC findings, achieving collaborative consensus on approved file versions.
Ghqc offered a viable solution by expanding on a credible, industry-standard tool, eliminating unfamiliar software burdens or extensive team training. The intuitive UI provided an enjoyable interface for users to have dynamic discussions. Manually recording important details for QC processes, such as version control information and file meta-data, is burdensome and prone to human error. The holistic approach of tracking QC within GitHub decreased this risk and provided a semi-automated solution for initiating, examining, resolving and recording the QC process.
Conclusions: Automation of QC is crucial for efficiency in review of files and accuracy in tracking version control and QC activities. In the era of AI, QC is an increasingly important final step in scientific workflows, providing manual validation by subject matter experts. By semi-automating QC, ghqc balances efficiency with making space for scientists to conscientiously assess code and output accuracy.
Citations: [1] R Core Team. (2023). R: A language and environment for statistical computing. R
Foundation for Statistical Computing. https://www.R-project.org/.
