Hi, I am attempting to preprocess 2000 samples on the HGU133 Plus 2 array. I have 16 batches of data, ranging in size between 50 to 200 samples. I was under the impression that using fRMA with the random effects model would be best (over median polishing or others) because it would help compensate for batch effects at the same time. However, I am limited in that I cannot read 200 CEL files into memory to do the processing. I was wondering: a) if there was a function or method similar to that of justRMA() for fRMA. I have read through all fRMA documentation and not found anything. b) what my best alternative would be. Right now I was thinking that I could run fRMA (median polishing) or SCAN.UPC (which one would be better?) on the samples individually and then use COMBAT to counter any batch effects. Please let me know if this would be an appropriate course of action. Thanks for your help! Sam [[alternative HTML version deleted]]

Sam, I would just use the default version of fRMA. Here you can preprocess the arrays individually or in small groups and then combine the preprocessed expression matrices. Both the default fRMA and the random effect version deal with batch effects in a similar way -- down weighting probes with high between-batch residual variance in our large training data set. The difference is the random effect method allows the probe effect in your data to differ slightly from the frozen probe effect. Given that you have at least one known batch variable, I would definitely examine the output of combat. You might also want to try SVA to see if there are any unknown batch variables present. Best, Matt On May 26, 2013 2:49 AM, "S.D. Chorlton" <chorltsd@mcmaster.ca> wrote: > Hi, > > I am attempting to preprocess 2000 samples on the HGU133 Plus 2 array. I > have 16 batches of data, ranging in size between 50 to 200 samples. > > I was under the impression that using fRMA with the random effects model > would be best (over median polishing or others) because it would help > compensate for batch effects at the same time. However, I am limited in > that I cannot read 200 CEL files into memory to do the processing. I was > wondering: > > a) if there was a function or method similar to that of justRMA() for fRMA. > I have read through all fRMA documentation and not found anything. > > b) what my best alternative would be. Right now I was thinking that I could > run fRMA (median polishing) or SCAN.UPC (which one would be better?) on the > samples individually and then use COMBAT to counter any batch effects. > Please let me know if this would be an appropriate course of action. > > Thanks for your help! > > Sam > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]