[BioC] limma and paired data

[Gordon Smyth]

At 03:09 AM 20/04/2004, Danielle Fletcher wrote:
>Hi,
>
>I am using limma to analyse a 2-colour microarray experiment. There are 2 
>treatments and 4 replicates in each of these groups. Each replicate is 
>paired to a replicate in the otehr treatment group. Each sample was 
>hybridised with a reference, so 8 slides in total.
>
>The targets file looks like this (hopefuly that will make it clearer):
>SlideNumber   Name   FileName   Cy3   Cy5
>1   1M   1.gpr   monolayer   ref
>2   1P   5b.gpr   pellet   ref
>3   2M   2.gpr   monolayer   ref
>4   2P   7.gpr   pellet   ref
>5   3M   3.gpr   monolayer   ref
>6   3P   6.gpr   pellet   ref
>7   4M   B.gpr   monolayer   ref
>8   4P   A.gpr   pellet   ref
>
>Initially my design matrix looked like this:
>
>    Sample-Ref Monolayer-Pellet
>1M           1          0
>1P           1          1
>2M           1          0
>2P           1          1
>3M           1          0
>3P           1          1
>4M           1          0
>4P           1          1
>
>but thinking about it again, i don't think this takes into account the 
>paired nature of the data. I am sure that the answer is probably a simple one,

There is no simple answer. There was a big discussion about this in 
Bioconductor very recently, please look at the list archives.

In the very latest versions of limma, there is a new argument 'block' in 
duplicateCorrelation() and lmFit() to handle a blocking structure like you 
describe. This feature is however very lightly documented so far and so is 
offered on a user beware basis. In your case, block=c(1,1,2,2,3,3,4,4).

Gordon

>  but I am not sure what the best solution is.  I would appreciate any advice.
>
>Thanks in advance
>
>Danielle