I would like to perform RMA style preprocessing for tiling arrays. Based on previous discussion (thanks Ben and Jim!), it seems that the functions rma.background.correct() and normalize.quantiles() from the preprocessCore package are the easiest way to accomplish this. First of all, I am having trouble getting the functions to work. The rma.background.correct() function seems to be returning the original values. The normalize.quantiles() function does something to the data, but the medians are not equal across arrays (which I would have expected from quantiles normalization). Secondly, I wanted to verify that to most closely match the RMA algorithm, I should (1) restrict to PM values (2) background correct, (3) normalize, then (4) log2 transform. Is this correct? The code (using Canine2 arrays for quicker experimentation) and sessionInfo are below. Ann ************************************** > library(affy) > library(preprocessCore) > AllArrays<-ReadAffy() > PM<-pm(AllArrays) > dim(PM) [1] 473162 15 > PM[1:10,1:5] RB1.CEL RB10.CEL RB11.CEL RB12.CEL RB13.CEL 571209 134 52 56 67 309 571210 151 63 78 85 587 571211 388 116 134 129 1396 571212 215 56 68 67 530 571213 83 41 42 47 131 571214 83 42 41 50 148 571215 107 43 42 63 194 571216 216 48 69 83 620 571217 254 61 80 79 876 571218 131 42 61 47 246 > BG<-rma.background.correct(PM) > dim(BG) [1] 473162 15 > BG[1:10,1:5] RB1.CEL RB10.CEL RB11.CEL RB12.CEL RB13.CEL 571209 134 52 56 67 309 571210 151 63 78 85 587 571211 388 116 134 129 1396 571212 215 56 68 67 530 571213 83 41 42 47 131 571214 83 42 41 50 148 571215 107 43 42 63 194 571216 216 48 69 83 620 571217 254 61 80 79 876 571218 131 42 61 47 246 > Norm<-normalize.quantiles(BG) > dim(Norm) [1] 473162 15 > Norm[1:10,1:5] [,1] [,2] [,3] [,4] [,5] [1,] 128.4667 143.40000 114.06667 332.7333 230.0000 [2,] 145.0000 202.66667 188.80000 560.2000 413.3333 [3,] 383.6000 480.93333 368.73333 1110.5000 936.6333 [4,] 208.6000 164.80000 155.20000 332.7333 375.8667 [5,] 79.6000 84.40000 65.26667 103.0667 108.0667 [6,] 79.6000 89.73333 62.00000 131.9333 120.0000 [7,] 102.3333 95.26667 65.26667 283.1333 151.9333 [8,] 209.4000 122.13333 158.80000 534.5333 435.6000 [9,] 247.5333 191.73333 195.80000 482.5333 602.5667 [10,] 125.7333 89.73333 131.33333 103.0667 187.9333 > apply(Norm,2,median) [1] 104.4667 106.1333 103.7333 103.0667 103.9333 104.2000 104.4667 [8] 104.4667 104.8000 104.0667 103.6667 103.6667 107.2000 106.0000 [15] 103.6667 > apply(PM,2,median) RB1.CEL RB10.CEL RB11.CEL RB12.CEL RB13.CEL RB14.CEL RB15.CEL 109 45 53 47 125 355 191 RB2.CEL RB3.CEL RB4.CEL RB5.CEL RB6.CEL RB7.CEL RB8.CEL 93 86 93 113 105 42 53 RB9.CEL 53 > sessionInfo() R version 2.7.1 (2008-06-23) i386-pc-mingw32 locale: LC_COLLATE=English_United States.1252;LC_CTYPE=English_United States.1252;LC_MONETARY=English_United States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252 attached base packages: [1] tools stats graphics grDevices utils datasets [7] methods base other attached packages: [1] canine2cdf_2.2.0 affy_1.18.2 preprocessCore_1.2.0 [4] affyio_1.8.0 Biobase_2.0.1

the issue with rma.background.correct is it failing to return the background corrected matrix when copy =TRUE (copy=FALSE will work). This is a genuine bug. It has been addressed in the devel-branch (http://fgc.lsi.umich.edu/cgi-bin/blosxom.cgi/preprocessCore/) at 1.3.1. Unfortunately, I did not patch the release branch with the fix (I should do so). Ben On Mon, 2008-07-21 at 11:13 -0600, Ann Hess wrote: > I would like to perform RMA style preprocessing for tiling arrays. Based > on previous discussion (thanks Ben and Jim!), it seems that the functions > rma.background.correct() and normalize.quantiles() from the > preprocessCore package are the easiest way to accomplish this. > > First of all, I am having trouble getting the functions to work. The > rma.background.correct() function seems to be returning the original > values. The normalize.quantiles() function does something to the data, > but the medians are not equal across arrays (which I would have expected > from quantiles normalization). > > Secondly, I wanted to verify that to most closely match the RMA algorithm, > I should (1) restrict to PM values (2) background correct, (3) normalize, > then (4) log2 transform. Is this correct? > > The code (using Canine2 arrays for quicker experimentation) and > sessionInfo are below. > > Ann > > ************************************** > > > library(affy) > > library(preprocessCore) > > AllArrays<-ReadAffy() > > PM<-pm(AllArrays) > > dim(PM) > [1] 473162 15 > > PM[1:10,1:5] > RB1.CEL RB10.CEL RB11.CEL RB12.CEL RB13.CEL > 571209 134 52 56 67 309 > 571210 151 63 78 85 587 > 571211 388 116 134 129 1396 > 571212 215 56 68 67 530 > 571213 83 41 42 47 131 > 571214 83 42 41 50 148 > 571215 107 43 42 63 194 > 571216 216 48 69 83 620 > 571217 254 61 80 79 876 > 571218 131 42 61 47 246 > > BG<-rma.background.correct(PM) > > dim(BG) > [1] 473162 15 > > BG[1:10,1:5] > RB1.CEL RB10.CEL RB11.CEL RB12.CEL RB13.CEL > 571209 134 52 56 67 309 > 571210 151 63 78 85 587 > 571211 388 116 134 129 1396 > 571212 215 56 68 67 530 > 571213 83 41 42 47 131 > 571214 83 42 41 50 148 > 571215 107 43 42 63 194 > 571216 216 48 69 83 620 > 571217 254 61 80 79 876 > 571218 131 42 61 47 246 > > > Norm<-normalize.quantiles(BG) > > dim(Norm) > [1] 473162 15 > > Norm[1:10,1:5] > [,1] [,2] [,3] [,4] [,5] > [1,] 128.4667 143.40000 114.06667 332.7333 230.0000 > [2,] 145.0000 202.66667 188.80000 560.2000 413.3333 > [3,] 383.6000 480.93333 368.73333 1110.5000 936.6333 > [4,] 208.6000 164.80000 155.20000 332.7333 375.8667 > [5,] 79.6000 84.40000 65.26667 103.0667 108.0667 > [6,] 79.6000 89.73333 62.00000 131.9333 120.0000 > [7,] 102.3333 95.26667 65.26667 283.1333 151.9333 > [8,] 209.4000 122.13333 158.80000 534.5333 435.6000 > [9,] 247.5333 191.73333 195.80000 482.5333 602.5667 > [10,] 125.7333 89.73333 131.33333 103.0667 187.9333 > > > apply(Norm,2,median) > [1] 104.4667 106.1333 103.7333 103.0667 103.9333 104.2000 104.4667 > [8] 104.4667 104.8000 104.0667 103.6667 103.6667 107.2000 106.0000 > [15] 103.6667 > > > apply(PM,2,median) > RB1.CEL RB10.CEL RB11.CEL RB12.CEL RB13.CEL RB14.CEL RB15.CEL > 109 45 53 47 125 355 191 > RB2.CEL RB3.CEL RB4.CEL RB5.CEL RB6.CEL RB7.CEL RB8.CEL > 93 86 93 113 105 42 53 > RB9.CEL > 53 > > > sessionInfo() > R version 2.7.1 (2008-06-23) > i386-pc-mingw32 > > locale: > LC_COLLATE=English_United States.1252;LC_CTYPE=English_United > States.1252;LC_MONETARY=English_United > States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252 > > attached base packages: > [1] tools stats graphics grDevices utils datasets > [7] methods base > > other attached packages: > [1] canine2cdf_2.2.0 affy_1.18.2 preprocessCore_1.2.0 > [4] affyio_1.8.0 Biobase_2.0.1 > > _______________________________________________ > 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