Hi there, I am helping my friend to do some analysis on his affy data. He has a simple 2x2 factorial experiment without any replication, i.e. only 4 chips. Because of this, I just perform MAS5, RMA and GCRMA on the raw data with the default settings, and compute the fold changes of individual factor comparison. I am suprised to see that there are many more genes which have >= 2 fold change in MAS5 compare to the other two algorithms. For example in one comparison, MAS5 gives 21477 genes >= 2 fold, while RMA gives 0 and GCRMA gives 4. While I do know that RMA gives less significant genes, I do not expect to see such a drastic difference. Especially when his experiment is a drug treatment on a cell culture which is known to change a lot of genes, I am expecting to see a lot of changes even in the fold change comparison. Is this discrepancy occured due to the lack of replication? Regards, Fai -- Yuk Fai Leung Department of Molecular and Cellular Biology Harvard University BL 2079, 16 Divinity Avenue Cambridge, MA 02138 Tel: 617-495-2599 Fax: 617-496-3321 email: yfleung@mcb.harvard.edu; yfleung@genomicshome.com URL: http://genomicshome.com

In BioConductor, the rma and gcrma expression measures are log2 transformed while MAS 5.0 is not. Could this be your reason ? Regards, Adai On Mon, 2004-09-13 at 15:48, YUK FAI LEUNG wrote: > Hi there, > > I am helping my friend to do some analysis on his affy data. He has a > simple 2x2 factorial experiment without any replication, i.e. only 4 > chips. Because of this, I just perform MAS5, RMA and GCRMA on the raw > data with the default settings, and compute the fold changes of > individual factor comparison. I am suprised to see that there are many > more genes which have >= 2 fold change in MAS5 compare to the other two > algorithms. For example in one comparison, MAS5 gives 21477 genes >= 2 > fold, while RMA gives 0 and GCRMA gives 4. > > While I do know that RMA gives less significant genes, I do not expect > to see such a drastic difference. Especially when his experiment is a > drug treatment on a cell culture which is known to change a lot of > genes, I am expecting to see a lot of changes even in the fold change > comparison. Is this discrepancy occured due to the lack of replication? > > Regards, > Fai

Dear Fai, I think justRMA() (If this is what you used) returns 2-log transformed values so that to >=2 fold changes correspond to (absolute difference)>=1 in that metric. Bests Martino On Monday, September 13, 2004, at 03:48 PM, YUK FAI LEUNG wrote: > Hi there, > > I am helping my friend to do some analysis on his affy data. He has a > simple 2x2 factorial experiment without any replication, i.e. only 4 > chips. Because of this, I just perform MAS5, RMA and GCRMA on the raw > data with the default settings, and compute the fold changes of > individual factor comparison. I am suprised to see that there are many > more genes which have >= 2 fold change in MAS5 compare to the other > two algorithms. For example in one comparison, MAS5 gives 21477 genes > >= 2 fold, while RMA gives 0 and GCRMA gives 4. > > While I do know that RMA gives less significant genes, I do not expect > to see such a drastic difference. Especially when his experiment is a > drug treatment on a cell culture which is known to change a lot of > genes, I am expecting to see a lot of changes even in the fold change > comparison. Is this discrepancy occured due to the lack of > replication? > > Regards, > Fai > -- > Yuk Fai Leung > Department of Molecular and Cellular Biology > Harvard University > BL 2079, 16 Divinity Avenue > Cambridge, MA 02138 > Tel: 617-495-2599 > Fax: 617-496-3321 > email: yfleung@mcb.harvard.edu; yfleung@genomicshome.com > URL: http://genomicshome.com > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > > --------------------------------------- Martino Barenco CoMPLEX 4, Stephenson Way London NW1 2HE Tel.: +44 20 7679 5088 Fax.: +44 20 7679 5052 (please specify "CoMPLEX") Email: m.barenco@ucl.ac.uk