Research reproducibility
Computational methods should be among the most reproducible methods in the life sciences; First Look: watch a web presentation of this work the environment and protocol for a computational method can be specified to a much greater degree, and with much greater precision, than a typical laboratory method. Unfortunately, many published computational results remain hard to reproduce, and methods are often not specified in sufficient detail to replicate or re-implement. I have proposed techniques for building reference environments11 Little bootstrapped open-source virtual environments for reproducing a single computational result. enabling researchers to reproduce computational results precisely, with minimal effort, and implemented these for a range of results in computational biology22 Lots of published examples here.
My current research in this theme extends this approach and set of tools to work across all major languages (R/Python/Java/MATLAB/Fortran/C) and platforms in systems and computational biology. The method produces reference environments across three different reproducibility platforms: as a virtual machine, as a container, and as a cloud environment, all from a single set of configuration scripts. Reference environments integrate readily with other reproducible research technologies, and can contain any type of language or computation.
The reference environment approach How do reference environments relate to other types of ‘reproducible research’? I discuss that here explicitly separates the core scientific findings of a piece of computational research from the software implementation in which they are embedded, and allows readers and reviewers to choose the most appropriate implementation type for their situation. Reference environments “But why not just use <currently popular reproducibility tool/website>?” I discuss that here also have many applications in more general reproducibility; providing standard environments for testing or benchmarking, or in teaching to accelerate students’ hands-on experience with tools and languages.
Watch a web presentation of this work here
Featured publications
Hurley, D. G., Budden, D. M., & Crampin, E. J. (2014). Virtual Reference Environments: a simple way to make research reproducible. Briefings in Bioinformatics, bbu043. doi:10.1093/bib/bbu043