[BioC] single color microarray analysis

[Sean Davis]

On Mar 3, 2005, at 1:27 AM, zhihua li wrote:

> Hi netters!
>
> I would like to know if there r any packages in bioconductor that can 
> perform single-color
> microarray analysis on the gene level instead of the probe level. 
> That's to say, I've had a matrix with each column representing the 
> expression values for a list of genes in a given condition and each 
> row represents the expression value of a gene across different 
> conditions. I wanna do all the analysis like normalization, 
> plotting....etc. I was at a loss what packages should I use?  The affy 
> package generally handles probe-level data - they require the colomns 
> to be signals for PM and MM probes, not the expression level of genes. 
> The limma package provides some tools to analysis my data, such as 
> normalizeQuantiles, etc. But limma generally handles two-color 
> microarray data, so there's not much can be done with my data. For 
> example, limma's plotting functions usually require both Red and Green 
> signal levels. I'm a bit surprised to see that there seems no packages 
> to handle single-color microarray data on the gene level. Could anyone 
> help me?  Thanks a lot!

Actually, limma will do single-channel analyses just fine.  If you have 
a matrix of expression values (genes in rows, samples in columns) 
appropriately normalized, you can happily use limma with that matrix.  
After all, the affy package ultimately produces just such a matrix.  
While the affy package does provide many bells and whistles specific to 
affy, there are many tools that you can use to evaluate your date.  
Look at limma's plotDensities function, which I think will let you plot 
single-channel data.  You can also use R's built-in plotting to do 
all-vs-all scatterplots to look at distribution issues.  You can 
threshold based on expression just using simple R subsetting.  And the 
full power of limma's linear models are not really platform-dependent, 
as I mentioned; they require only a matrix of numbers.  So, I the tools 
are there.  Read the single-channel sections of the limma manual, but 
where an eset (or expression set, in affy parlance) is used, you can 
simply use your numeric matrix of expression values.

Hope this helps.
Sean