[BioC] Question about design in limma

[Gordon K Smyth]

Hi January,

To compare KO to WT, you need to do a separate channel anlaysis, using 
intraspotCorrelation() and lmscFit().  This will allow you to make any 
comparison you like, as if you had 40 single-channel arrays, but taking 
into account the two-colour structure.  See the section on this in the 
limma User's Guide.

Best wishes
Gordon

> Date: Mon, 17 Jan 2011 01:50:35 +0100
> From: January Weiner <january.weiner at mpiib-berlin.mpg.de>
> To: BioC <bioconductor at stat.math.ethz.ch>
> Subject: [BioC] Question about design in limma
> Content-Type: text/plain; charset=ISO-8859-1
>
> Hello,
>
> I've been presented with the following design:
>
>   Tissue strain  Cy5  Cy3
> 1   tis1     KO   d0  d01
> 2   tis1     KO  d01   d0
> 3   tis1     KO   d0  d03
> 4   tis1     WT  d03   d0
> 5   tis1     WT   d0  d01
> 6   tis1     KO  d03   d0
> 7   tis1     WT  d01   d0
> 8   tis1     WT   d0  d03
> 9   tis2     KO   d0  d01
> 10  tis2     KO  d01   d0
> 11  tis2     KO   d0  d02
> 12  tis2     KO  d02   d0
> 13  tis2     WT   d0  d01
> 14  tis2     WT  d01   d0
> 15  tis2     WT   d0  d02
> 16  tis2     WT  d02   d0
> 17  tis3     WT   d0  d03
> 18  tis3     WT  d03   d0
> 19  tis3     KO  d03   d0
> 20  tis3     KO   d0  d03
>
> In summary: there are three different tissues. There are two strains
> and four measurement points for the experiment (d0, d1, d2, d3), with
> d0 as reference. The two-channel arrays have been loaded with RNA
> samples such that the "day 0" from the respective tissue is the
> reference. There is a dye swap present.
>
> I am looking for differences in the strains -- which genes are
> regulated in a different manner between WT and KO. How should I do it
> best? obviously, simply comparing WT with KO is *not* the way to go,
> since change in d04 (as compared with d0) can be very different from
> response in d02, and also response is very variable between tissues.
> However, I am looking for genes that are, for example, differentially
> expressed in the KO, but not changing in the WT.
>
> One way of solving that would be splitting the data sets, fitting the
> model separately to the subsets, creating lists for subsets (e.g. list
> of genes diff. expressed in KO, list of DE genes in WT) and comparing
> the subsets. It works, but I don't like it.
>
> Best regards,
>
> j.
>
> -- 
> -------- Dr. January Weiner 3 --------------------------------------
> Max Planck Institute for Infection Biology
> Charit?platz 1
> D-10117 Berlin, Germany
> Web?? : www.mpiib-berlin.mpg.de
> Tel? ?? : +49-30-28460514

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}