[BioC] difference in results using different edgeR package versions

[Steve Lianoglou]

Hi,

On Fri, Aug 2, 2013 at 7:32 AM, Fernando Biase <fernandobiase at gmail.com> wrote:
> Dear all,
> I did a RNA-seq analysis two years ago (2011), using edgeR package (I do
> not remember the version). I was asked to analyze the data again, and now I
> am using edgeR_3.2.4.
>
> The comparison is between two groups. The "backbone" of the code was very
> similar (it is pasted bellow), and I just changed it to adapt the package
> modifications. It turned out that the number of differentially expressed
> genes (FDR<0.05) in the second analysis (78) was a smaller than the number
> of DEG obtained in the first analysis (119).
>
> I understand that two years is a lot of time in such fast evolving field,
> and I suppose that the newer package version has more reliable analysis
> than older version. Nonetheless, has anyone experienced something similar?
> Is there a clear reason for this (other than I am making a mistake)?

This topic comes up every now and again, and I believe it is usually
due to some changes to the defaults for the prior.df (or prior.n)
params somewhere.

I'm sure that searching the archives (in the footer of each bioc mail
is the link to search the archives via gmane) will help to shed more
light on this, and perhaps this email thread would be a good place to
start:

http://thread.gmane.org/gmane.science.biology.informatics.conductor/46826

HTH,
-steve

-- 
Steve Lianoglou
Computational Biologist
Bioinformatics and Computational Biology
Genentech