Dear Ana, Your idea sounds reasonable in principle to me. It seems reasonable to me that summarized values would have somewhat less tight correlation than R/G values on the same spot, so I wouldn't be worried about an intraspot correlation of 0.5. Best wishes Gordon > Date: Mon, 9 Aug 2010 19:02:02 +0200 > From: Ana Conesa <aconesa at="" cipf.es=""> > To: "bioconductor at stat.math.ethz.ch" <bioconductor at="" stat.math.ethz.ch=""> > Subject: [BioC] [limma] Separate Channel Analysis of 2 color data with > summarization > > > Dear List, > > I am analyzing some yeast Nimblegen arrays following the separate channel > analysis of 2 color data of the limma package. After within and between > arrays normalization steps I did a summarization (medianpolish of M and A > values for 12 spots/gene) step to reduce the number of features for > differential expression analysis. My idea was then to calculate > intraspotcorrelation and use the lmscFit function. My question is whether it > should be OK to compute the intra spot correlation and the linear model using > these summarized values. One result that calls my attention is that the > intraspotcorrelation computed on the summarized values has a consensus value > of 0.517. From other posts in the list, it has been mentioned that intraspot > correlation should be above 0.8, so I am wondering if this approach is > conceptually correct. > > Thanks for any suggestion on this topic > > Regards > > -- > Ana Conesa > Bioinformatics and Genomics Department > Centro de Investigaciones Principe Felipe > Avda. Autopista Saler 16, > 46012 Valencia, Spain > Phone: +34 96 328 96 80 > Fax: +34 96 328 97 01 > http://bioinfo.cipf.es/aconesa > http://www.blast2go.org ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}