All AWS batch jobs run in a docker image, which is great, and we should use community-based docker images as much as possible. Most standalone tools already have them via bioconda/biocontainers.
However, for R packages, it probably makes as much sense to maintain our own with the packages we need. As this will not run interactively for the most part, we may be able to base it off an image without RStudio installed, which would potentially save some space, but flexibility for development outside of AWS might outweigh this concern.
All AWS batch jobs run in a docker image, which is great, and we should use community-based docker images as much as possible. Most standalone tools already have them via bioconda/biocontainers.
However, for R packages, it probably makes as much sense to maintain our own with the packages we need. As this will not run interactively for the most part, we may be able to base it off an image without RStudio installed, which would potentially save some space, but flexibility for development outside of AWS might outweigh this concern.