I was looking at my cuffdif gene differential expression data, and I noticed that I can't exactly determine where the number in column 8 and 9 (labeled as value_1 and value_2) actually comes from. It is not the average of the cufflinks gene expression FPKM data from my three replicates, but it is close enough that I think it must be related. What is the process by which Cuffdif creates its combined replicate value, that is not a simple average? It must have to do with how the sum of the reads changes when you merge the replicates, but I couldn't easily find exactly what is happening, so any help would be appreciated.
A description of the output fields is here: http://cole-trapnell-lab.github.io/cufflinks/cuffdiff/#differential-expression-tests
Other sections of the website site above, including the link outs to the paper, describe the calculations in detail (they are too complex to simplify here). Cufflinks and Cuffdiff are not expected to produce the same FPKM values as the fragment counts change once data is merged.
Prior Q&A about this and related subjects can be seen in the sidebar here (Similiar posts >>) and other places on the web, including the Google group for the tool, also linked from the same site above. Galaxy RNA-seq tutorials can be found here: https://galaxyproject.org/learn/
Hope this helps! Jen, Galaxy team