Git for Reproducibility in Scientific Research

Together with the SCQM research team we introduced tools and established processes to facilitating collaboration among researchers and increase transparency and reproducibility.

«Reproducibility is the hallmark of good science.» http://www.scfbm.org/content/8/1/7/

Git (Version Control System)

Git is a decentralized and distributed version control system (VCS). It's widely used in the software industry to maintain code repositories.

LaTeX (Typesetting)

LaTeX is a high-quality typesetting system. LaTeX is the de facto standard for the publication of scientific documents.

latex-git-log can be used to type­set the complete Git version history as table written in LaTeX. The typeset output (PDF document) includes the Git revisions (short SHA-1 hash). http://www.ctan.org/tex-archive/support/latex-git-log

SQL Research Database (Snapshots)

All data of the registry is mirrored to a separate SQL database (research database). During the transfer process data transformations take place to bring the data into a tabular structure and to anonymize personal subject data. To ensure full reproducibility of SQL queries we create periodic snapshots of the research database. A snapshot is a complete immutable SQL database with a consistent naming convention (YYYY-MM-DD: e.g. 2013-12-29)

Sweave (R in LaTeX)

Sweave combines R codes and LaTeX documents for reproducing results. It makes it easy to embed R code into LaTeX.

Project Templates

For each new research project a template based Git repository is  created. The template contains a common folder layout, scripts and LaTeX templates. The deployment of new repositories is managed with Puppet.

Secure Shell Access (SSH)

Access control to the Git research repositories and to the SQL server with the research snapshot database is managed with SSH public/private keys.

GitWeb (Git Server)

Git repositories on a Git server have a URL and could be opened to the public by submitting links to the Git repository alongside manuscripts sharing them with reviewers. To lower technical barrier (installing and maintaining a Git server) web-based hosting service for Git repositories such as GitHub or Bitbucket could be used for the collaboration as well.


About SCQM
The SCQM Foundation runs a scientific platform for prospective long-term studies on inflammatory rheumatic diseases