[BioC] down-weighting control probes in limma (with Agilent arrays)?

[Daniel Aaen Hansen]

Dear List,

I have a question regarding what weighting function to use when analyzing Agilent data with limma. My question is whether or not it makes sense to down-weight the control probes?

I have defined a weighting function:

wtfun.Agilent = function(x) {
	okPopnOutlier    = x[,"rIsFeatPopnOL"]==0 & x[,"gIsFeatPopnOL"]==0
	okNonUnifOutlier = x[,"rIsFeatNonUnifOL"]==0 & x[,"gIsFeatNonUnifOL"]==0
	okControlType    = x[,"ControlType"]==0
	as.numeric(okPopnOutlier & okNonUnifOutlier & okControlType)
}

And I would then use read.maimages() to read the data:

RG = read.maimages(targets, source="agilent", wt.fun=wtfun.Agilent)

This line assigns weight 0 (FALSE) to control probes and weight 1 (TRUE) to non-control probes:

okControlType    = x[,"ControlType"]==0

As far as I can see, the normalization methods don't make explicit use of the control probes and hence I would think it would be a good idea to down-weight them. Any opinions on this?

Are there any way to make use of the controls?

Best,
Daniel