[BioC] limma question

[Gordon K Smyth]

> Hi
> I am working with a chip which contains "empty" spots, i.e. nothing has
> been
> spotted there, "landing lights", and actual spots "cDNA". I would like to
> exclude "empty" and "landing light" spots from my dataset before
> normalisation.
>
> So, this works fine:
> library(limma)
> targets=readTargets(...)
> spottypes=readSpotTypes(...)
> files = c(targets$FileNameCy3, targets$FileNameCy5)
> numrows = length(files)/2
> dim(files) = c(numrows,2)
> rawdata = read.imagene(files, path=...)
> g2=data.frame(ID=rawdata$genes$"Gene ID",Name=rawdata$genes$"Gene ID")
> status = controlStatus(spottypes,g2)
> normdata = normalizeWithinArrays(rawdata)
>
> ... and I tried this to exclude the "empty" spots and "landing lights":
>
> raw2=rawdata[status=="cDNA",]
> normdata = normalizeWithinArrays(raw2,method="none")
>
> ... which works fine, but with
>
> normdata = normalizeWithinArrays(raw2)

Use instead

w <- as.numeric(status=="cDNA")
normdata <- normalizeWithinArrays(rawdata,weights=w)

Remember that limma expects to get complete arrays if you wish to use a
print-tip based normalization method.  Therefore keep all the spots in
your data object but give zero weight to the spots you don't want to have
any influence on the normalization.

Gordon

> I get this
>>Error: subscript out of bounds
>
> Any suggestions?
>
> Thanks
> Thomas
>
> _________________________
> Thomas Schlitt
> British Antarctic Survey
> th_schlitt at gmx.de