Dear Bioconductor users, I have 12 Affymetrix arrays with a factorial design 2 x 2. I've been using the package Affy, GCRMA and Limma to analize my data. Now I have a list of diferential expressed genes for each factor and I was wondering if some of these genes have an expression level in all the arrays near to the background for removing them from my final list (it's an "eyeballometric" method and the cut-off is arbitrary, I know). So, I've been observing the expression levels of some genes in Affy and in GCRMA and I've observed some differences. I expected some differences, but they are bigger than I thought. For exemple, gene 1370943_at: Expression levels for MAS5.0 (not log2 transformed) Array names: DNT1.CEL DNT2.CEL ...until 12 Expression levels: 4.2 7.3 ------------->"eyeballometrically" I had removed this gene (there isn't any other gene in my list with lower expression levels) Expression levels for GCRMA (log2 transformed) Array names: DNT1.CEL DNT2.CEL ....until 12 Expression levels: 20.68613483 27.21562194 ->"eyeballometrically" I hadn't removed this gene (there are other genes in my list with values between 6 and 7) Moreover, the expression levels of the Perfect Match and Miss Match are: Expression levels for pm: >pm(data,"1370943_at") DNT1.CEL DNT2.CEL ......until 12 1370943_at1 503.8 434.3 1370943_at2 1084.5 857.8 1370943_at3 413.3 378.5 1370943_at4 102.8 119.5 1370943_at5 339.3 325.0 1370943_at6 380.5 327.3 1370943_at7 154.5 117.5 1370943_at8 197.5 195.0 1370943_at9 70.8 77.5 1370943_at10 55.5 62.0 1370943_at11 55.0 72.0 Expression levels for mm: >mm(data,"1370943_at") DNT1.CEL DNT2.CEL ....until 12 1370943_at1 645.5 535.3 1370943_at2 1250.5 1119.0 1370943_at3 415.0 499.5 1370943_at4 144.5 127.8 1370943_at5 281.8 291.0 1370943_at6 448.8 444.0 1370943_at7 196.3 174.3 1370943_at8 127.0 131.5 1370943_at9 71.3 75.0 1370943_at10 61.8 73.0 1370943_at11 61.3 81.3 Therefore, I suppose I should remove this gene from my list, should I? Is there any way to stablish a cut-off for the lower expressing genes? The "eyeballometric" system is too subjective. Thank you very much for you time and help. Yours faithfully, Jordi Altirriba PhD student Hospital Clinic - Barcelona - Spain

> Expression levels for GCRMA (log2 transformed) > Array names: DNT1.CEL DNT2.CEL ....until 12 > Expression levels: 20.68613483 27.21562194 ->"eyeballometrically" > I hadn't removed this gene (there are other genes in my list with values > between 6 and 7) It is surpring to me that GCRMA would give such high log2 transformed expression for this gene. What does the distribution of expression values look like? > > Moreover, the expression levels of the Perfect Match and Miss Match are: > > Expression levels for pm: > >pm(data,"1370943_at") > DNT1.CEL DNT2.CEL ......until 12 > 1370943_at1 503.8 434.3 > 1370943_at2 1084.5 857.8 > 1370943_at3 413.3 378.5 > 1370943_at4 102.8 119.5 > 1370943_at5 339.3 325.0 > 1370943_at6 380.5 327.3 > 1370943_at7 154.5 117.5 > 1370943_at8 197.5 195.0 > 1370943_at9 70.8 77.5 > 1370943_at10 55.5 62.0 > 1370943_at11 55.0 72.0 > > Expression levels for mm: > >mm(data,"1370943_at") > DNT1.CEL DNT2.CEL ....until 12 > 1370943_at1 645.5 535.3 > 1370943_at2 1250.5 1119.0 > 1370943_at3 415.0 499.5 > 1370943_at4 144.5 127.8 > 1370943_at5 281.8 291.0 > 1370943_at6 448.8 444.0 > 1370943_at7 196.3 174.3 > 1370943_at8 127.0 131.5 > 1370943_at9 71.3 75.0 > 1370943_at10 61.8 73.0 > 1370943_at11 61.3 81.3 > > Therefore, I suppose I should remove this gene from my list, should I? Is > there any way to stablish a cut-off for the lower expressing genes? The > "eyeballometric" system is too subjective. > > Thank you very much for you time and help. > Yours faithfully, > > Jordi Altirriba > PhD student > Hospital Clinic - Barcelona - Spain > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >

Dear Zhijin Wu, Yes, you are rigth. I've done this stupid mistake. I'm sorry. Thanks, Jordi Altirriba PhD Student Hospital Clinic - Barcelona - Spain >From: Zhijin Wu <zwu@jhsph.edu> >To: Jordi Altirriba Guti?rrez <altirriba@hotmail.com> >Subject: Re: [BioC] Expression levels in GCRMA and MAS5.0 and cut-off >Date: Wed, 15 Sep 2004 11:13:12 -0400 (EDT) > >You can see that the range of gcrma expression values is about (0,17). >I don't understand how you could have a gene with expression value >22 or even 27 from gcrma. It doesnt look consistent with the graph. > >is it possible that those number were transformed back to the linear >scale? gu?a telef?nica de QDQ. http://qdq.msn.es/msn.cfm