[BioC] limma and nested factors

[Arne.Muller at aventis.com]

Hi,

thanks for reply. I guess you mean is to use the rank as given by topTable
(by p-value) for one contrast for a comparison with another contrast? Since
the rank is already given (by topTable) would I've to use a pearson
correlation?

What I've just tried is a simple pearson (the data look not too far from
normal) and spearman correlation separate for up and downregulated genes.

> cor.time.spearman
               down          up
NEWvsOLD  0.4461673 0.371619276
NEWvsPRG -0.1675682 0.003799389
OLDvsPRG -0.1500851 0.057734972

> cor.time.pearson
                 down         up
NEWvsOLD  0.723402369 0.49514365
NEWvsPRG -0.035844726 0.01057829
OLDvsPRG  0.005358755 0.04535618

Given are the "estimates". The correlation is based on the coefficients in
the eBayes fit from limma. Only those coefficients were included for which
either one or the other (e.g. NEW or OLD) provides a coefficient >0 (up) with
p<=0.01, or <0 for down. For each correlation between 500 and 1000 genes are
included.

This is not too far from what I expect, since OLD and NEW have been generated
using quite similar experimental protocols. It also tells that these batches
don't have that much in common ...

	thanks for your suggestion,

	Arne

--
Arne Muller, Ph.D.
Toxicogenomics, Aventis Pharma
arne dot muller domain=aventis com

> -----Original Message-----
> From: Stephen Henderson [mailto:s.henderson at ucl.ac.uk]
> Sent: 08 April 2004 15:06
> To: 'Gordon Smyth '; Muller, Arne PH/FR
> Cc: 'bioconductor at stat.math.ethz.ch '
> Subject: RE: [BioC] limma and nested factors
> 
> 
> how about the rank spearman correlation of all the toptable p-values?
> 
> >There seems nothing wrong with your limma analysis. Focusing 
> only on the
> >most extreme differentially expressed genes as you do with a 
> toptable will 
> >of course exaggerate the differences between the three labs.
> 
> 
> 
> **********************************************************************
> This email and any files transmitted with it are confidential and
> intended solely for the use of the individual or entity to whom they
> are addressed. If you have received this email in error please notify
> the system manager (it.support at wibr.ucl.ac.uk). All files are 
> scanned for viruses.
> **********************************************************************
> 
>