This is in fact what we do here in Montreal. We find that the beadstudio software is kind of useless. Using version 3.x, we export the raw probe-level data, and surrogate replace all any values that are below the background controls using next largest log value +1;ex if the mean background controls are 59 we replace with 64+1 or 65 (2^n+1), if its 110 it would be 128+1 or 129. This eliminates the issue of negative numbers. Background subtraction just makes the analysis more difficult (especially doing local background subtraction on some platforms) Then we perform the quantile normalization using the functions in Limma. Peter > -----Original Message----- > From: Mark Dunning [mailto:md392 at cam.ac.uk] > Sent: Wednesday, June 20, 2007 1:45 PM > To: daneel jordan > Cc: bioconductor at stat.math.ethz.ch > Subject: Re: [BioC] Illumina Treatment > > Hi Daneel, > > To my knowledge there are is no suitable benchmark dataset > available through Bioconductor for comparing methods. > > In our experience, we find the method of normalisation > implemented in BeadStudio is not helpful if you want to > analyse data on the log-scale using the tools available in > Bioconductor. This is because of the huge number of NA values > that are produced and large increase in variability at low > intensities. > > I would recommend exporting non-normalised data fromn > BeadStudio and then normalising using traditional methods > such as quantile after a log2 transformation or using the VST > method of lumi. > > Regards, > > Mark > > > On Tue, 2007-06-19 at 11:42 +0200, daneel jordan wrote: > > Dear Bioconductor user, > > > > > > I need general help in illumina data treatments. I > previously see that > > there is Lumi packages to treat data, but I didn't see any > benchmarch > > to compare data normalization provided by lumi, and the ones from > > BeadStudio. (if you have any can you provide any link?); with no > > benchmarks I would like to keep the analysis as much standart as > > possible, using Rank-invariant from Bead Studio. > Unfortunately, as you > > may know for sure, this treatment create negative values, > because of > > bg subtraction, after log transformation, I have NaN value, > when I run > > a simple ttest analysis for each line of the array it can > happen that it complains for NAN values. what do you suggest? > > add to the all chip a minim value in order to eliminate all > possible > > negative values? > > threshold the data for negative values? > > is there any way to tell the function t.test to not > consider NaN fields? > > > > Daneel > > > > [[alternative HTML version deleted]] > > > > _______________________________________________ > > 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 > > _______________________________________________ > 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 > >