Hi, all, I need some suggestions =) I'd like to use package metaArray to perform meta-analysis. One of the tools of this package is integrative correlation analysis (proposed Parmigiani et al. in 2004), it allows to evaluat concordance of genes expression between studies. The main function is intcor() which allows to get integrative correlations, then I can filter genes (reject genes that is uncorrelated between studies), argument of intcor() function is merged object, combination of exprSet (package affy). I have different samples in exprSet (case and control). Should I separate case-samples from control-samples before analysis? Or intcor() function can distinguish samples (maybe it gets information from phenoData slots of exprSets...) Thanks a lot, Olga

> Message: 4 > Date: Sun, 29 Apr 2007 02:25:18 +0700 > From: " ????? ??????? " <kamushkina at="" gmail.com=""> > Subject: [BioC] Help with intcor() function > To: bioconductor at stat.math.ethz.ch > Message-ID: > <d1419f550704281225r76c93ac7y6561334a8297783d at="" mail.gmail.com=""> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > Hi, all, > > I need some suggestions =) > > I'd like to use package metaArray to perform meta-analysis. > One of the tools of this package is integrative correlation analysis > (proposed Parmigiani et al. in 2004), it allows to evaluat concordance > of genes expression between studies. > The main function is intcor() which allows to get integrative > correlations, then I can filter genes (reject genes that is > uncorrelated between studies), argument of intcor() function is > merged object, combination of exprSet (package affy). > I have different samples in exprSet (case and control). Should I > separate case-samples from control-samples before analysis? > Or intcor() function can distinguish samples (maybe it gets > information from phenoData slots of exprSets...) > > Thanks a lot, > Olga Hi Olga, intCor() is part of the MergeMaid package. Integrative correlation is an unsupervised method for obtaining a filtered set of features that are more reproducible, so it does not use the case-control variables in phenoData and you should not separate the cases from the controls. The vignette and help page may a few examples you can follow: http://www.bioconductor.org/packages/2.0/bioc/html/MergeMaid.html Best regards, Rob