An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20050919/ ba684dde/attachment.pl

Hi Bruz, I can't speak to the values for GCRMA, but my suspicion would be that there are slight differences in rounding between the implementations, explaining the differences at the lowest intensity for RMA. However, I can not attest to the correctness or otherwise of the ArrayAssist implementation of the algorithm. For all intents and purposes the BioConductor implementation as given by the rma() function should be the canonical version (justRMA calls all the same underlying routines so it will return identical expression values). Ben On Mon, 19 Sep 2005, Bruz Marzolf wrote: > Hi all, > > > > Has anyone tried validating RMA and GC-RMA in Stratagene's ArrayAssist > Lite software against the Bioconductor implementations? ArrayAssist Lite > is appealing in that it's free and capable of performing these methods > on massive numbers of CEL files on a desktop computer with limited > memory-it just takes a really long time. > > > > I typically use the Bioconductor implementations of RMA/GC-RMA, so I > wanted to make sure that if I generated data in ArrayAssist Lite it > would be comparable. Despite using what appear to be the same parameters > for these methods with both software packages, the results are not > coming out the same. For RMA, the Pearson correlation of the log(signal) > coming out of the two methods is R=0.997, with slight disagreement in > values at low signal values. For GC-RMA, I'm getting R=0.98, and in the > low signal range ArrayAssist Lite is showing much higher signal for some > probe sets. Here are the software versions and parameters: > > > > Bioconductor: > > justRMA() function from affy version 1.5.8, default parameters > > justGCRMA() fuction from gcrma version 1.1.3, default parameters > > > > ArrayAssist Lite 3.3, default parameters for both functions > > > > I'd appreciate hearing about experiences anyone has had with this sort > of comparison, or any feedback on where I may be going wrong. > > > > Thanks, > > Bruz > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >