[BioC] Limma and time-course data

[michael watson (IAH-C)]

To complete my original mail, I guess I was referring to this approach:

http://bioinf.wehi.edu.au/marray/jsm2005/lab5/lab5.html
 

-----Original Message-----
From: Sean Davis [mailto:sdavis2 at mail.nih.gov] 
Sent: 28 February 2006 13:22
To: michael watson (IAH-C); Bioconductor
Subject: Re: [BioC] Limma and time-course data




On 2/28/06 7:30 AM, "michael watson (IAH-C)"
<michael.watson at bbsrc.ac.uk>
wrote:

> Hi
> 
> Googling the list shows this up to be a rather hot topic, but I just 
> wanted to ask a few more questions.
> 
> Firstly, it seems the plan for tackling time course data within limma 
> is to treat each time-point/treatment combination as a coefficient to 
> be estimated.  Thus, to ask "which genes are changing over time", one 
> must fit contrasts that compare every single time point to every other

> time point, pairwise, and look for any gene that is significant in one

> or more of those comparisons.  Is that correct?

I would say that this is only one of several ways of analyzing
time-course data, and perhaps not the best one for all situations.  In
fact, sometimes I have the best solution to be simple filtering of genes
followed by clustering and display, but I think the "correct" solution
depends on the experimental design (numbers of time points, for example)
and goals (for example, it doesn't help a biologist to have 2000 genes
in a list if the goal is to find 10 transcription factors that seem to
be affected at any time point).  

For limma, you could use decideTests, for example, to give you some
sense of genes that are changing in the experiment.  Or you could filter
for those genes that are changing at  the "maximal" time point and then
show those genes for all the timepoints on a heatmap--this will allow
the biologist to quickly focus on genes of interest.

Sean