Hello Bioconductor Users, Sorry to bother you, but I have a couple of quick questions on the affy package that I have been unable to answer by reading through the textual description of affy available on the bioconductor website and the recent papers on the subject. Iin general, am very impressed with the noise reduction using affy and truly appreciate the Bioconductor open source concept and packages available. I am trying to compare directly our own homebrewed data normalization within Splus (splining of affyIDs (post MAS summing of probes)) with the rma output. Question 1: Are the values output by rma to eset (and written to tab delimited file with write.exprs) the logbase2 of the "actual" affyID values, or are they the affyID values themselves. This is obviously important for comparing rma to our old methods, and for estimates of "accurate" fold change estimates- It seems to me that rma is much more precise in that much of the noise between replicates, especially at the low end, seeems to have been reduced in rma, but I am struggling with how to estimate back to accurate fold change estimates. Question 2: Is there an easy way to write out the PM values to tab delimited file post background correction and normalization (to assess individual probe sets prior to median polish). I am trying to work out reasons for differences between our old method and rma for individual affy IDs. Sincerely, John Luckey C John Luckey, MD PhD Post Doctoral Fellow - Benoist-Mathis Lab Joslin Diabetes Center One Joslin Place, Rm. 474 Boston, MA 02215 phone: (617) 264-2783 fax: (617) 264-2744 e-mail: john.luckey@joslin.harvard.edu

> > Question 1: Are the values output by rma to eset (and written to tab delimited file with write.exprs) the logbase2 of the "actual" affyID values, or are they the affyID values themselves. This is obviously important for comparing rma to our old methods, and for estimates of "accurate" fold change estimates- It seems to me that rma is much more precise in that much of the noise between replicates, especially at the low end, seeems to have been reduced in rma, but I am struggling with how to estimate back to accurate fold change estimates. > The values given by the rma() function are log base 2. Working in log base 2 you could take differences and then take 2^(difference) to get fold changes. Or you could just take 2^(rma expression value) and take ratios if you prefer. You might find that RMA estimated fold changes are attenuated compared to other methods. That is that the estimated fold change is smaller than that computed using other methods. But the noise reduction at the low end is pretty dramatic. > Question 2: Is there an easy way to write out the PM values to tab delimited file post background correction and normalization (to assess individual probe sets prior to median polish). I am trying to work out reasons for differences between our old method and rma for individual affy IDs. If you manually background correct and normalize (as individual steps), doing something like data <- ReadAffy() data.bg <- bg.correct.rma(data) data.norm.bg <- normalize(data,method="quantiles") in each case pmdata.bg) and pmdata.norm.bg) will give pm matrices after preprocessing. Investigate the function write.table() for outputting the data to tab delimited files. thanks, Ben