[BioC] Vsn transformation repeatability

[Paolo Innocenti]

Wolfgang Huber wrote:
> 1. You say "number of ... differentially expressed genes for a
> particular FDR changes with each run". By how much? The variation should
> be tiny (negligible?), otherwise I would be worried about the stability
> of your method to compute differential expressedness, and FDR.

I use the simplest possible approach in limma (one factor with 2 levels) 
following the userguide. The problem seems to be that the distribution 
of adjusted p.values is rather odd: instead of having a peak at 0, I 
have a clear peak between 0.05 and 0.1, so that a minimum difference in 
the normalization have a rather big effect on the number of my 
differentially expressed genes (I fixed the cutoff at 0.1).

The difference is the order of 5% of d.e. probesets found (range between 
1700 and 1900 on a total of 18000).

> Please see the code example below for how to deactivate the subsampling.

Deactivating subsampling fix the problem!
Thank you very much!

Best,
paolo

-- 
Paolo Innocenti
Department of Animal Ecology, EBC
Uppsala University
Norbyvägen 18D
75236 Uppsala, Sweden