[BioC] pre-processing Agilent data to remove a bad sample

[Francois Pepin]

Hi Stephen,

You might want to analyse your chips like single-channel ones, try to
correct for the dye effect and then go on normally. At this point I
doubt you'd want to be looking at log ratios.

I've never worked with the marrayRaw objects so I'm not sure how you'd
handle it there, using ExpressionSets or Limma objects I'd set each
sample as if it was its own array.

Have anyone else tried to run two samples on a single chip like that? I
know people have used them as single-color arrays, but I don't remember
seeing it used as two single-color arrays and I'm curious how much
effect that the competitive hybridization would have on the signal
intensity.

Francois

On Tue, 2007-02-13 at 10:10 +0200, Stephen Rudd wrote:
> Dear bioconductor colleagues
> 
> I am fighting with a support project using two colour Agilent arrays. As
> expected with the typical academic microarray study, there are too many time
> points and not enough replicates.
> 
> Within the problem study I have 19 arrays, containing 38 samples. The data looks
> pretty good throughout but a single sample is a dramatic outlier within
> correspondence analysis and clustering. I therefore wish to remove this sample
> from the analysis (but I don't have much experience with Agilent).
> 
> Data has been built into an marrayRaw object, and I have removed the sample from
> the object, but the loess normalisation will not proceed; I suspect that paired
> data is required?
> 
> My question is therefore how could (should) I remove this data point
> representing the signal from one channel of an array. This data should be
> removed earlier within the analysis because there are some rather subtle DEG
> effects later on in the time series. I don't wish to remove the whole array
> since the paired data is of considerable value (through lack of replicates!).
> Would it make more sense to remove this data channel from the result marrayNorm
> object (the first normalisation is within array).
> 
> Any comments, suggestions or otherwise would be gratefully received
> 
> Thanks
> 
> Stephen
> 
> --
> Dr Stephen Rudd
> Adjunct professor of plant genomics
> Senior specialist, bioinformatics
> A schizophrenic coexistence between pharma and academia
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
>