Philip, The issue I am raising here is that I have seen a few situations where there is a large clear artifact on a chip, reflected in the NUSE and RLE statistics as being of poorer quality, using the standard procedure, but when called with the GCRMA background the artifacts are gone and there is little or nothing in the NUSE/RLE statistics to suggest anything wrong. To me it just seems too good to be true to believe that the GCRMA background, or any other for that matter, can magically fix all such artifacts. However, the above is based on looking at only a couple of datasets and I am happy to be proved wrong. Since we are discussing the topic of assessing data quality, as I do from time to time, I encourage people to check out the gallery: http://www.stat.berkeley.edu/users/bolstad/PLMImageGallery/index.html And I am always looking for new submissions (this is a hint to anybody who might be interested in exploring the above issue in a more serious way). Thanks, Ben ------------ B. M. Bolstad bolstad at stat.berkeley.edu http://www.stat.berkeley.edu/~bolstad On Tue, 18 Oct 2005, Groot, Philip de wrote: > Hello Ben, > > Quality assessment is not my only concern. I am performing subsequent > analyses after GC-RMA normalization (using e.g. the Limma package) and > looking at fitPLM images using the same background correction is a good > thing. But perhaps it is a bit unclear to me what you exactly want to > say. > > And yes, I did use the option: background.method="GCRMA". However, I was > not aware of the threestep functionality. Thanks for the tip! > > Regards, > > Philip > > -----Original Message----- > From: Ben Bolstad [mailto:bolstad at stat.Berkeley.EDU] > Sent: Tuesday, October 18, 2005 4:00 PM > To: Groot, Philip de > Cc: bioconductor at stat.math.ethz.ch > Subject: Re: [BioC] R-2.2 and BioConductor 1.7: changes in GC-RMA and > fitPLMprocedures > > Hi, > > It is good to hear that the results agree. The function bg.correct.gcrma > was introduced into the affyPLM package at the time of the previous BioC > release (1.6) as an aid to allow you to use the GC-RMA background with > the > fitPLM, threestep, threestepPLM etc (ie you could supply > background.method="GCRMA" as an argument to any of these functions). In > this latest release the GCRMA now includes bg.adjust.gcrma which does > essentially the same correction (as you have discovered) but with a > slightly different set of arguments. These differences will be removed > in > the future. > > As an aside, I would be a little wary about combining the GCRMA > background > with fitPLM when your only interest is quality assessment. > > Ben > > On Tue, 18 Oct 2005, Groot, Philip de wrote: > > > Hello, > > > > Just a small additional note. I just found out that both commands: > > bg.adjust.gcrma() and bg.correct.gcrma() return exactly the same > > results, thus solving my problem. I was confused by two different > > function calls that, in the end, do the same thing. I apologize for > any > > inconvenience. > > > > Regards, > > > > Dr. Philip de Groot > > Wageningen University > > > > > > -----Original Message----- > > From: Groot, Philip de > > Sent: Tuesday, October 18, 2005 10:16 AM > > To: bioconductor at stat.math.ethz.ch > > Subject: [BioC] R-2.2 and BioConductor 1.7: changes in GC-RMA and > > fitPLMprocedures > > > > Hello, > > > > > > > > I submit this message to the mailing list because I think that more > > people encounter the same problem. In general, when performing quality > > control calculations, performing fitPLM calculations (affyPLM package) > > is a good idea. Normalizing your data is also a good idea, so (in our > > situation) the same calculations are performed two times: fitPLM > > executes an GC-RMA background correction and the same thing is done > when > > applying the GC-RMA normalization. This was for me a reason to combine > > both calculations in a separate script in BioC 1.6, in which the > GC-RMA > > background correction is performed, used as input for the fitPLM > > procedure, and the further normalization steps are executed afterwards > > (this required digging in both scripts and perform the separate > required > > steps in a new script). > > > > > > > > Unfortunately, in BioC 1.7 things has changed significantly in such a > > way that it is more difficult to do the above things: all GC-RMA > > calculations (except the summarization) are performed in > > "bg.adjust.gcrma" while the fitPLM GC-RMA background correction is > > performed in "bg.correct.gcrma" (without an option to pass the > > previously calculated affinities to this function and to get the > > in-between GC-RMA values back). > > > > > > > > So, my question (request) is straightforward: are people working on a > > better integration of these two scripts, so that the end result of > > fitPLM can be used for the further GC-RMA normalization procedure? > This > > is not only more efficient, but also saves a considerable amount of > > computation time! > > > > > > > > Kind regards, > > > > > > > > Dr. Philip de Groot > > > > Wageningen University > > > > > > [[alternative HTML version deleted]] > > > > _______________________________________________ > > Bioconductor mailing list > > Bioconductor at stat.math.ethz.ch > > https://stat.ethz.ch/mailman/listinfo/bioconductor > > >