[BioC] Spot weights

[Gordon Smyth]

Hi Jason,

At 08:10 PM 2/06/2003, Jason Skelton wrote:
>Hi
>
>could someone give me an example of how to use the spot quality (name.W)
>option from marrayRaw, I have a marrayRaw object with name.W taken from a
>flags column from genepix (gpr) files but I'm not sure how to use it to
>discard data in subsequent normalisations.  The gpr files have either P
>(pass), M (marginal) F (fail) OR 0,-50 and -100 respectively.

I'd encourage you to persevere with marrayRaw. I am sure that what you want 
to do can be done with maNormMain.

>Can you do this in R as well as bioconductor ?
>for data that isn't marrayRaw

Well, an alternative is to use the limma package, which is designed to make 
it easy to use spot quality weights. If 'files' is a character vector 
containing the names of the GenPix gpr files, you want to give zero weight 
to the non-passing spots, and your gal file is in your working directory, 
then you can type

library(limma)
gal <- readGAL()
RG <- read.maimages(files, source="genepix", wt.fun=wtflags(0))
MA <- normalizeWithinArrays(RG, getLayout(gal))

This will perform print-tip group loess normalization giving zero weight to 
all spots which GenePix gives a negative flag. I think I'd probably use 
wt.fun=wtflags(0.1) or similar which gives 10% weight to the non-passing 
spots, just to keep them in the normalizing regressions.

Cheers
Gordon

>cheers
>
>Jason