[BioC] Different results Limma vs LimmaGUI?

[Gordon Smyth]

>[BioC] Different results Limma vs LimmaGUI?
>J.delasHeras at ed.ac.uk J.delasHeras at ed.ac.uk
>Tue Apr 11 16:00:16 CEST 2006
>
>Hi,
>
>I have been using LimmaGUI for a while, and I am starting to use Limma
>directly, and I was comparing the results I get with both methods, to
>make sure I am doing things right.
>
>The lists of genes I am getting are roughly the same, but the orders do
>change a bit, and so do the B values (that I use to rank the genes).
>
>Is this expected, perhaps because different limma versions do
>calculations in slightly different ways? or is that an indication that
>I am not doing the analysis in the same way?
>
>For LimmaGUI, I use the latest (July 2005), version 1.3.9 (which uses
>limma 2.0.2) on R 2.1.1
>Limma, I am using version 2.2.0 with R 2.2.0 (both latest)

limma 2.2.0 was the version of limma at the time of last Bioconductor 
release, but it is not the latest version of limma. Actually limma 
2.4.13 is available for R 2.2.0. Please read Section 2.2 
(Installation) of the Limma User's Guide for why this is so.

>In one experiment (two slides only) I get a B max of 6.33 with limma,
>and 6.92 with LimmaGUI. Another 3.13 with limma, and 2.84 with
>LimmaGUI. Then I add another 3 slides (5 total now) and the B max
>increases to 13.0 in both limma and limmaGUI. So, the differences are
>not big... is this due to slight changes in the way successive versions
>of limma do certain calculations?

>The analysis included:
>- bkg correction: substract
>- within arrays normalisation: loess print-tip
>- between arrays normalisation: scale
>- least squares (for linear model) in limmaGUI (default in limma)

There is no reason that I know of that should cause limma 2.02 and 
limma 2.2.0 to give different results for the B-statistic. There was 
a small change made in limma 2.4.0, and another in limma 2.4.13.

You should probably become familiar with the limma change log. Type 
changeLog(n=280) at the command line, and you'll see an entry for 
limma 2.4.0 dated 11 Nov 2005.

Best wishes
Gordon

>Thanks for your help,
>
>Jose
>
>--
>Dr. Jose I. de las Heras                      Email: J.delasHeras at ed.ac.uk
>The Wellcome Trust Centre for Cell Biology    Phone: +44 (0)131 6513374
>Institute for Cell & Molecular Biology        Fax:   +44 (0)131 6507360
>Swann Building, Mayfield Road
>University of Edinburgh
>Edinburgh EH9 3JR
>UK