[BioC] Limma final gene expression report

[michael watson (IAH-C)]

I can see where it may be handy though - e.g. for input into clustering.
Really, at that stage, most people would prefer one gene - one value.

There's no way I would average over the values for different oligos
though, even if they were for the same gene!

In limma, in "usersguide.pdf", section 14 "Within Array Replicate Spots"
deals with the issue :-)

Mick

-----Original Message-----
From: Sean Davis [mailto:sdavis2 at mail.nih.gov] 
Sent: 10 May 2005 12:11
To: Ankit Pal
Cc: michael watson (IAH-C); bioconductor at stat.math.ethz.ch
Subject: Re: [BioC] Limma final gene expression report




On May 10, 2005, at 6:54 AM, Ankit Pal wrote:

> There are different oliogos of the same gene and
> replicates with the same sequence present on the
> array.
> Splice variants have been mapped to the same Accession
> number but I'm working on that to separate out the
> same by naming them differently.
> -Ankit
>

So, for the different oligos/splice variant spots, averaging is 
probably NOT a good idea in most situations--they could be measuring 
quite different things.  When there is a discrepancy, it MAY be due to 
array issues (quality of one spot, for instance), but it could also be 
a true finding; determining which is which often requires some human 
intervention.

Perhaps others will weigh in on the issue, but I don't think ad-hoc 
averaging of truly duplicated spots really serves a purpose, either.  
Unless your array design allows you to treat the duplicated spots at 
the level of the analysis (lmFit, in limma) and not post-processing, I 
don't think you benefit much from averaging (i.e., you don't have more 
"confidence" in a gene that has been averaged).

Sean