Hi list, I kind of remember that someone mentioned this before, but wouldn't find it in the archive. I normalized 9 arrays ( 3 replicates for each of 3 conditions) used GCRMA. Surprisingly, I found that for one gene (or more) that the normalized expression values are identical for 2 conditions (total 6 arrays) as > light.gcrma.exprs[22497,] 189.02.CEL 189.03.CEL 189.04.CEL LER.02.CEL LER.03.CEL LER.04.CEL 194.02.CEL 3.368551 3.368551 3.368551 3.368551 3.368551 3.368551 3.763626 194.03.CEL 194.04.CEL 3.368551 3.349966 Note the first 6 values are identical My question is: is it possible? and how does this happen? Since the expression is pretty low (might not express at all), does GCRMA do something like "flooring". I checked RMA normalized data, it has identical values for two replicates for only one condition. If I use gcrma with fast=FALSE, I got > light.gcrma.exprs[22497,] 189.02 189.03 189.04 LER.02 LER.03 LER.04 194.02 194.03 2.769276 2.706325 2.658057 2.837865 2.880565 2.781661 2.186927 2.370975 194.04 2.554258 Although this gene itself shouldn't enter next step analysis, I just want to make sure that nothing wrong with GCRMA algorithm. R: 2.1.0 patched GCRMA: 1.1.3 Window XP R code > light.gcrma=gcrma(data.light) ##where data.light is affy batch object. (or light.gcrma=gcrma(data.light, fast=FALSE) ) > light.gcrma.exprs=exprs(light.gcrma) Thanks in advance! Fangxin -------------------- Fangxin Hong Ph.D. Plant Biology Laboratory The Salk Institute 10010 N. Torrey Pines Rd. La Jolla, CA 92037 E-mail: fhong at salk.edu (Phone): 858-453-4100 ext 1105

There are two steps where it's possible probe level data are forced to be the same: in "fast=TRUE" version of GCRMA, intensities way below the expected background level are forced to a same lower limit. In quantile normalization if one probe gets the same rank across arrays, even if its absolute observed intensity is not identical on all arrays, the normalization converts the intensities to the same number. At gene level, one does not usually observe many identical measurements unless the above happens for all or many probes in a probeset. But that is still possible especially when the gene is probably not expressed. best, Jean > > My question is: is it possible? and how does this happen? Since the > expression is pretty low (might not express at all), does GCRMA do > something like "flooring". I checked RMA normalized data, it has identical > values for two replicates for only one condition. If I use gcrma with > fast=FALSE, I got > > light.gcrma.exprs[22497,] > 189.02 189.03 189.04 LER.02 LER.03 LER.04 194.02 194.03 > 2.769276 2.706325 2.658057 2.837865 2.880565 2.781661 2.186927 2.370975 > 194.04 > 2.554258 > > > > Although this gene itself shouldn't enter next step analysis, I just want > to make sure that nothing wrong with GCRMA algorithm. > > R: 2.1.0 patched > GCRMA: 1.1.3 > Window XP > > R code > > light.gcrma=gcrma(data.light) ##where data.light is affy batch object. > (or light.gcrma=gcrma(data.light, fast=FALSE) ) > > light.gcrma.exprs=exprs(light.gcrma) > > Thanks in advance! > > > Fangxin > > -------------------- > Fangxin Hong Ph.D. > Plant Biology Laboratory > The Salk Institute > 10010 N. Torrey Pines Rd. > La Jolla, CA 92037 > E-mail: fhong at salk.edu > (Phone): 858-453-4100 ext 1105 > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >

RMA and GCRMA use quantile normalization step at individual probe level. It is possible to get identical values after summarising these values. I recently had constant value across 6 arrays for a particular gene. But I guess any kind of oddity might happen with 50,000+ genes. Regards, Adai On Thu, 2005-08-11 at 11:28 -0700, fhong at salk.edu wrote: > Hi list, > > I kind of remember that someone mentioned this before, but wouldn't find > it in the archive. > > I normalized 9 arrays ( 3 replicates for each of 3 conditions) used GCRMA. > Surprisingly, I found that for one gene (or more) that the normalized > expression values are identical for 2 conditions (total 6 arrays) as > > light.gcrma.exprs[22497,] > 189.02.CEL 189.03.CEL 189.04.CEL LER.02.CEL LER.03.CEL LER.04.CEL 194.02.CEL > 3.368551 3.368551 3.368551 3.368551 3.368551 3.368551 3.763626 > 194.03.CEL 194.04.CEL > 3.368551 3.349966 > Note the first 6 values are identical > > > My question is: is it possible? and how does this happen? Since the > expression is pretty low (might not express at all), does GCRMA do > something like "flooring". I checked RMA normalized data, it has identical > values for two replicates for only one condition. If I use gcrma with > fast=FALSE, I got > > light.gcrma.exprs[22497,] > 189.02 189.03 189.04 LER.02 LER.03 LER.04 194.02 194.03 > 2.769276 2.706325 2.658057 2.837865 2.880565 2.781661 2.186927 2.370975 > 194.04 > 2.554258 > > > > Although this gene itself shouldn't enter next step analysis, I just want > to make sure that nothing wrong with GCRMA algorithm. > > R: 2.1.0 patched > GCRMA: 1.1.3 > Window XP > > R code > > light.gcrma=gcrma(data.light) ##where data.light is affy batch object. > (or light.gcrma=gcrma(data.light, fast=FALSE) ) > > light.gcrma.exprs=exprs(light.gcrma) > > Thanks in advance! > > > Fangxin > > -------------------- > Fangxin Hong Ph.D. > Plant Biology Laboratory > The Salk Institute > 10010 N. Torrey Pines Rd. > La Jolla, CA 92037 > E-mail: fhong at salk.edu > (Phone): 858-453-4100 ext 1105 > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >