How to: Migrate and reproduce an analysis elsewhere

Q: I've just finished a project. How can my collaborators reproduce the results on their machines?

A: In order to replicate an analysis, (1) the annotation data files, (2) the raw data files, and (3) the R script need to be available. By transferring (1)-(3) to a new file system and rerunning the script will reproduce the results exactly, given the same setup of packages and version of R. This should also be true if this is done between different operating systems. Here is what you need to do in order to transfer the analysis to you collaborators:

1. Identify and copy/zip the content of the raw data set directory/directories, e.g. rawData/HapMap270,CEU/Mapping250K_Nsp/ and rawData/HapMap270,CEU/Mapping250K_Sty/.
2. Identify and copy/zip the content of the annotation chip type directory/directories, e.g. annotationData/chipTypes/Mapping250K_Nsp/ and annotationData/chipTypes/Mapping250K_Sty/. If there are auxiliary annotation files, you may want to copy only the ones used in the analysis.
3. Identify and copy/zip all R scripts.

Note that all of the above data is available from the current working directory (either directly or via file/directory links). Furthermore, the R scripts should not have to be modified, but should work "as is" regardless of location and operating system. This is because the aroma.* framework is designed to work without explicit file paths (if you find that you have to specify a path, you are probably doing something wrong).

It is good practice to always do the above and validate the reproducible of any analysis completed, before passing it on to collaborators or submitting to a journal. When doing this one often find minor discrepancies such as not using the annotation files that one believe one uses and so on. It is also a good way to archive the analysis done.