Dear Naomi, > After single channel normalization, I usually use MA.RG to transform back to > R and G. ?It has worked remarkably well. Yes, this is also my usual procedure. However, as you noticed, this does not help us with getting rid of the intra-array correlation. I was hoping that there might me a standard procedure to correct for this. While normally one does not really need the R & G channels, I have particular situations where I would like to have them, like the abovementioned meta-study. In that particular case it would be extremely important to get rid of any correlations between the channels in one array. Kind regards, January However, I retain the array effect > in the model. > > Regards, > Nasomi Altman > > > > At 03:55 AM 2/8/2012, January Weiner wrote: >> >> Dear all, >> >> first, I would like to thank all who answered my questions in the past. >> >> I am attempting a meta-analysis of several microarray studies, with >> limma as my working horse. I plan to throw all the microarrays >> together, creating one large data set with one of the factors in the >> analysis being the data set. Preliminary analyses with a limited >> number of studies are encouraging; on one hand, I am able to reproduce >> the results of the single studies, while at the same time finding >> robust differences between the studies (and their respective cohorts). >> >> However, I hit a wall with one of these studies, which involved >> two-color Agilent chips, not with a common reference, but each chip >> corresponding to two different individuals from two experimental >> groups. For some of the analyses that I plan I need separate >> intensities for each experimental group -- fold changes won't cut it >> (for example, in case of machine learning in which I use the >> intensities to construct a model for predicting the group >> assignments). >> >> I ?tried to directly use the R and G channels, and the results are >> actually quite good. Of course, this is not an optimal approach. >> Normally, when faced with two-color arrays and a complex experimental >> design I use intraspotCorrelation and lmscFit. >> >> Question: is there a way to use the results of intraspotCorrelation to >> correct the R and G channels? >> >> Kind regards, >> j. >> >> -- >> -------- Dr. January Weiner 3 -------------------------------------- >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor at r-project.org >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor > > > > -- -------- Dr. January Weiner 3 -------------------------------------- Max Planck Institute for Infection Biology Charit?platz 1 D-10117 Berlin, Germany Web?? : www.mpiib-berlin.mpg.de Tel? ?? : +49-30-28460514 Fax ? ?: +49-30-28450505