in my lab we are thinking about restyling our pipeline management system and would like to exploit as much as possible the wealth of annotations available in the AnnotationData Packages.
Our system is based on a set of available scripts for common operations and makefiles
and aims at being as most language agnostic as possible - right now we have a bunch of 'annotation dirs' organized with versions and/or download dates in pathnames which stores genomes, gene coords and so on in flat tab delimited files.
Ideally we would like to switch to annotation packages without generating with them those files but by performing the task at hand directly in the working projects directories (i.e. getting the sequences corresponding to a given set of coords or converting some gene ids, etc) with a set of ad hoc R scripts.
In this scenario the problem of version management arises: it's not only necessary to write somewhere the used version of R/Bioc/packages but also ideally to be able (possibly in a user transparent way, i.e. setting a makefile variable) to re-run the same steps (that are makefile rules) with the same original versions.
I was wondering if someone else organizes pipelines in a manner similar to ours and has faced the same issues. I am thinking about linking our makefile management system with docker images with the right package versions but I am a bit worried about the disk requirements of such an approach.
I did some googling but haven't found anything really fitting our scenario - I hope that I did not miss anything obvious. I saw this enabling reproducible research & R package management & install.package.version & BiocLiteold post but I believe that it is a bit outdated.
Thank you very much for your help, I am in general curious about pipeline management inside the Bioconductor community, maybe everyone is using knitr and that's it :) - I use it for some solo project but still haven't tried it out for big ones and for the kind of undergraduate student that we work with starting with that would be troublesome unfortunately.