Hi! We have data from an experiment conducted to compare three different methods (1,2 and 3) for amplifying RNA. RNA from two tissues (A and B) were amplified with the three different methods and labeled. The same dye was assigned to the same tissue in all hybridizations (no dye- swap). Three slides were hybridized within each method. Slide Method Cy3 Cy5 1 1 A B 2 1 A B 3 1 A B 4 2 A B 5 2 A B 6 2 A B 7 3 A B 8 3 A B 9 3 A B The data from each slide is background corrected and loess normalized within each slide followed by quantile normalization between slides. We want to test whether the three different methods reveal the same genes to be differentially expressed between tissue A and B - and whether two of the methods resemble each other more than the third. How should the design/contrast matrix look like? ------------------------------------------------------------------- Jakob Hedegaard Danish Institute of Agricultural Sciences Department of Genetics and Biotechnology Molecular Genetics and System Biology Building K25 Research Centre Foulum P.O. box 50 DK-8830 Tjele Denmark Tel: (+45)89991363 Fax: (+45)89991300 [[alternative HTML version deleted]]

This is an example of a "three group" problem. See the section called "Several Groups" in the limma User's Guide, version 1.8.18 or later. You'll find this in the Bioc developmental repository or at http://bioinf.wehi.edu.au/limma You want the "group-means" parametrization for the design matrix, which tells you which genes are DE for each method, and pairwise differences contrasts, which tells you how different the methods are. Gordon >Date: Tue, 8 Feb 2005 17:05:28 +0100 >From: "Jakob Hedegaard" <jakob.hedegaard@agrsci.dk> >Subject: [BioC] LIMMA: analysis of direct design.... >To: <bioconductor@stat.math.ethz.ch> > >Hi! > > > >We have data from an experiment conducted to compare three different >methods (1,2 and 3) for amplifying RNA. RNA from two tissues (A and B) >were amplified with the three different methods and labeled. The same >dye was assigned to the same tissue in all hybridizations (no dye- swap). >Three slides were hybridized within each method. > > > >Slide Method Cy3 Cy5 > >1 1 A B > >2 1 A B > >3 1 A B > >4 2 A B > >5 2 A B > >6 2 A B > >7 3 A B > >8 3 A B > >9 3 A B > > > >The data from each slide is background corrected and loess normalized >within each slide followed by quantile normalization between slides. > >We want to test whether the three different methods reveal the same >genes to be differentially expressed between tissue A and B - and >whether two of the methods resemble each other more than the third. > > > > > >How should the design/contrast matrix look like? > > > > > > > >------------------------------------------------------------------- > >Jakob Hedegaard > >Danish Institute of Agricultural Sciences > >Department of Genetics and Biotechnology > >Molecular Genetics and System Biology > >Building K25 > >Research Centre Foulum > >P.O. box 50 > >DK-8830 Tjele > >Denmark > >Tel: (+45)89991363 > >Fax: (+45)89991300