[BioC] doing some math on an array of PM probes

[James W. MacDonald]

Stuart Brown wrote:
> 
> I would like to normalize probe signals per gene on a single Affy array 
> by dividing
> each pm probe by the mean of all pm probes on that array. Never mind 
> that this is a
> stupid thing to do, I have an idea...
> 
> The command  pm(affybatch)
> gives a single probe value in each row and adds a numerical suffix to 
> the gene name so
> that each probe has a unique gene name.
> 
> Instead,  I'd like to extract the pm probes into a matrix that has a 
> single Affy gene name for a row header and the signal values for all of 
> the pm probes for that gene in the following columns of that row. Some 
> genes will have different numbers of columns since they have different 
> numbers of
> pm probes. [If I could just get a table in this shape, I could actually 
> do all the rest of the work
> in Excel.]

You will have a hard time making a matrix in R that has different column 
lengths (recycling rules will make the matrix rectangular). You will be 
much better off using a list and the lapply() function.

mn <- mean(pm(abatch)) ## where abatch is the name of your AffyBatch
norm.pm <- lapply(pm(abatch, LISTRUE=TRUE), function(x) x / mn)

This will give you a list of all PM probes normalized by the overall mean.

HTH,

Jim

> 
> Then I'd like to take the mean of all the probe signal values for each 
> gene and write that into a new array that has just 2 rows - genename 
> (which serves as the index) and mean-pm for that gene.
> 
> Finally, I'd like to replace the values of each pm probe signal with the 
> ratio of that probe signal divided by  the mean of all pm probes for 
> that gene.
> 
> There must be an easy way to do this, but I'm new to Bioconductor.
> 
> Thanks for any and all helpful suggestions
> -Stuart Brown
> 
> 


-- 
James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109