Hello ! I am comparing the transcriptome of two yeast strains, A and B. For each strain we have 5 biological replicates. The strains are compared to a reference on two-color arrays. For each biological replicate, we used two slides, in dye-swap. So we have 20 slides : slide Cy3 Cy5 1 A1 ref 2 ref A1 3 A2 ref 4 ref A2 5 A3 ref 6 ref A3 7 A4 ref 8 ref A4 9 A5 ref 10 ref A5 11 B1 ref 12 ref B1 13 B2 ref 14 ref B2 15 B3 ref 16 ref B3 17 B4 ref 18 ref B4 19 B5 ref 20 ref B5 In addition, the spots are duplicated on each slide. So we have biological replicates, technical replicates (dye-swaps) and spot duplication. We are interested in genes differentially expressed between both strains. I carefully read the Limma User's Guide, but it seems that it's not possible to handle duplicated spots and technical replicates simultaneously. In addition I'm not sure how to design my contrast matrix for this combination of biological and technical replicates in a common reference design... I would appreciate any help, thank you ! Sylvie Pinloche - INRA

Hi Sylvie, Here is some advice I got from Gordon Smyth about 2.5 years ago: My approach to designs like this, at least as a start, is to try duplicateCorrelation() on each of the levels separately to get an idea of the strength of the correlation at each level. Very often, some of the levels are so weak that they can be ignored. So, first ignore the dye-swaps and set up your design matrix as if you have 10 biological replicates of each strain, then use duplicateCorrelation() on the duplicate spots. Second, ignore the duplicate spots, use the same design matrix but use duplicateCorrelation() with the dye-swap pairs as the blocking variable. If the correlation on the dye-swap pairs is not substantially negative (~ -0.2??), then you could just ignore it and use duplicateCorrelation() on the duplicate spots. On the other hand, if the correlation of the duplicate spots is extremely high, you could just average the values and then use duplicateCorrelation() on the dye-swaps. This latter option is my preference; see ?avedups for an easy way to average the duplicate spots. HTH, Jenny At 10:35 AM 11/12/2008, sylvie pinloche wrote: >Hello ! > >I am comparing the transcriptome of two yeast strains, A and B. For >each strain we have 5 biological replicates. >The strains are compared to a reference on two-color arrays. For >each biological replicate, we used two slides, in dye-swap. >So we have 20 slides : > >slide Cy3 Cy5 >1 A1 ref >2 ref A1 >3 A2 ref >4 ref A2 >5 A3 ref >6 ref A3 >7 A4 ref >8 ref A4 >9 A5 ref >10 ref A5 >11 B1 ref >12 ref B1 >13 B2 ref >14 ref B2 >15 B3 ref >16 ref B3 >17 B4 ref >18 ref B4 >19 B5 ref >20 ref B5 > >In addition, the spots are duplicated on each slide. >So we have biological replicates, technical replicates (dye-swaps) >and spot duplication. We are interested in genes differentially >expressed between both strains. >I carefully read the Limma User's Guide, but it seems that it's not >possible to handle duplicated spots and technical replicates >simultaneously. In addition I'm not sure how to design my contrast >matrix for this combination of biological and technical replicates >in a common reference design... >I would appreciate any help, thank you ! > >Sylvie Pinloche - INRA > >_______________________________________________ >Bioconductor mailing list >Bioconductor@stat.math.ethz.ch >https://stat.ethz.ch/mailman/listinfo/bioconductor >Search the archives: >http://news.gmane.org/gmane.science.biology.informatics.conductor Jenny Drnevich, Ph.D. Functional Genomics Bioinformatics Specialist W.M. Keck Center for Comparative and Functional Genomics Roy J. Carver Biotechnology Center University of Illinois, Urbana-Champaign 330 ERML 1201 W. Gregory Dr. Urbana, IL 61801 USA ph: 217-244-7355 fax: 217-265-5066 e-mail: drnevich@illinois.edu [[alternative HTML version deleted]]