Hello everyone, I am a statistician at the Ontario Cancer Institute. Recently, I had to analyse an affymetrix dataset. The data have been normalized and the level expressions were calculated using GC RMA as implemented in R. My role is to analyse the expression levels (which were calculated using GC RMA) using SAM and other statistical techniques. The expression levels which were given to me (calculated with GC RMA) are very large: all are >1. I wonder if this is what one would expect. The analyst assures me that a log transformation was also applied. I am not sure how to interpret this numbers. What would be the levels of: 'not expressed', 'over expressed', or 'under expressed'? For example I know that the expression levels (not logged) obtained using cDNA microarray can be interpreted (ideally)as: 1=no expression, >2 over expression and <0.5 under expression. I appreciate your help. Thank you. Melania Pintilie Biostatistics Department Ontario Cancer Institute University of Health Network/Princess Margaret Hospital Toronto, M5G 2M9 Canada Tel: (416) 946-4501 ext. 4886 Fax: (416) 946-2048

Melania Affymetrix chips are "one-color" experiments, so the expression you refer to is NOT the log-ratio of expression of experimental compared to reference that you see in "two-color" cDNA experiments. Affy expression is an "absolute expression, in the sense that high numbers mean more molecules and low numbers mean few molecules, although the numbers can only be compared within gene and cannot be compared between genes (i.e. if one gene has a higher number than another, that does not necessarily mean that the first gene had more molecules). Susan Pintilie, Melania wrote: >Hello everyone, > >I am a statistician at the Ontario Cancer Institute. Recently, I had to >analyse an affymetrix dataset. The data have been normalized and the level >expressions were calculated using GC RMA as implemented in R. >My role is to analyse the expression levels (which were calculated using GC >RMA) using SAM and other statistical techniques. >The expression levels which were given to me (calculated with GC RMA) are >very large: all are >1. I wonder if this is what one would expect. The >analyst assures me that a log transformation was also applied. > >I am not sure how to interpret this numbers. What would be the levels of: >'not expressed', 'over expressed', or 'under expressed'? > >For example I know that the expression levels (not logged) obtained using >cDNA microarray can be interpreted (ideally)as: 1=no expression, >2 over >expression and <0.5 under expression. > >I appreciate your help. >Thank you. > > >Melania Pintilie >Biostatistics Department >Ontario Cancer Institute >University of Health Network/Princess Margaret Hospital >Toronto, M5G 2M9 >Canada >Tel: (416) 946-4501 ext. 4886 >Fax: (416) 946-2048 > >_______________________________________________ >Bioconductor mailing list >Bioconductor at stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor > > -- ---------------------------------->8================================= Susan Galloway Hilsenbeck, PhD........mailto:sgh at breastcenter.tmc.edu Breast Center at Baylor College of Medicine... MaBell: (713) 798-1627 One Baylor Plaza, MS BCM600................... Fax: (713) 798-1642 Houston, TX 77030

On Jun 10, 2005, at 8:26 AM, Pintilie, Melania wrote: > > For example I know that the expression levels (not logged) obtained > using > cDNA microarray can be interpreted (ideally)as: 1=no expression, >2 > over > expression and <0.5 under expression. > Your best bet is to use statistical methods to determine expression in 1 group of experiments relative to another. Look at the vignette for multtest and the user guide for Limma (http://bioinf.wehi.edu.au/limma/usersguide.pdf) for examples. Sean

Pintilie, Melania wrote: > Hello everyone, > > I am a statistician at the Ontario Cancer Institute. Recently, I had to > analyse an affymetrix dataset. The data have been normalized and the level > expressions were calculated using GC RMA as implemented in R. > My role is to analyse the expression levels (which were calculated using GC > RMA) using SAM and other statistical techniques. > The expression levels which were given to me (calculated with GC RMA) are > very large: all are >1. I wonder if this is what one would expect. The > analyst assures me that a log transformation was also applied. > > I am not sure how to interpret this numbers. What would be the levels of: > 'not expressed', 'over expressed', or 'under expressed'? A recommendation I've recieved from statisticans in my University (Tel-Aviv University in Israel) is to use MAS Absent/Present marks as a filtering guide, if possible. Remove all those genes who were marked Absent or Marginal.