Hello, I would like to follow up on a discussion of how to best mimic dchip analysis in BioConductor. Web documentation for the dchip PM-only algorithm states that normalization is based on the ORIGINAL PM and MM data, and the background correction is performed AFTER normalization for the PM-only model. (http://biosun1.harvard.edu/complab/dchip/pm%20only.htm) Based on the order of operations used by expresso (1-background correction, 2-normalization, 3- probe specific background correction, 4- summary into expression measure), it does not seem possible to Normalize the data (using invariant set algorithm) and then background correct (using mas algorithm) to exactly mimic the dchip implementation. So, it seems unclear as to whether or bg.correct=FALSE or bgcorrect.method="mas" is more appropriate. Is there a way to change the order of operations for expresso? Ann On Wed, 29 Sep 2004, Ben Bolstad wrote: > Actually I believe that it is a slightly modified version of the MAS > background that is used in dChip. They use a 10 by 10 grid rather than 4 > by 4. > > Ben > > > On Wed, 2004-09-29 at 08:37, Lizhe Xu wrote: >> Hi, >> The Li and Wong method uses the .cel file generated by MAS. What I believed the bg.correct=MAS in order to exactly mimic it. >> I don't know why to use bg.correct=FALSE in your analysis. >> >> Thanks. >> >> L. >> >> Ann, >> >> A more exact way to mimic the Li and Wong method with the command >> expresso is: >> expresso(affybatch, normalize.method="invariantset", >> bg.correct=FALSE, pmcorrect.method="subtractmm", >> summary.method="liwong") >> >> As you say, Li and Wong state their model as: >> PM_ij - v_j = Theta_i*phi_j + e >> with v_j a background term. >> >> (One can note that is currently disputed >> whether MM_ij is an appropriate background term or not.) >> >> >> remark: If you use this often you can make your own wrapper >> liandwong <- function(abatch, ...) { >> expresso(affybatch, normalize.method="invariantset", >> bg.correct=FALSE, pmcorrect.method="subtractmm", >> summary.method="liwong", ...) >> } >> >> Hoping this helps, >> >> >> L. >> >> >> Ann Hess wrote: >>> I was wondering if there was a way to obtain background (or non- specific >>> binding) corrected values for the Li Wong PMonly method. Li and Wong >>> state their model as PM_ij = v_j + Theta_i*phi_j + e. Is it possible to >>> find the value PM_ij - v_j? >>> >>> Is the command: >>> expresso(affybatch.example,normalize.method="invariantset", >>> bg.correct=FALSE,pmcorrect.method="pmonly",summary.method="liwong") >>> making this "background correction"? >>> The background term cancels for the PM-MM model, but not for the PM only >>> model. >>> >>> Thanks >>> >>> Ann >>> >>> _______________________________________________ >>> Bioconductor mailing list >>> Bioconductor@stat.math.ethz.ch >>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>> >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@stat.math.ethz.ch >> https://stat.ethz.ch/mailman/listinfo/bioconductor > -- > Ben Bolstad <bolstad@stat.berkeley.edu> > http://www.stat.berkeley.edu/~bolstad > >

Hi. I'm using affyPLM models to check for some array quality. Here you can see the plots of the residuals www.med.unibs.it/dip/dip_SBB/sez_statist/calza/bioinfo So, afaik some of those chip look a bit suspicious to, i.e. LptMOE8 and LptMOE9 and also LptMOE11 and LptMOE12. How should I interpret those "stains"? Are this array that bad? Is it reasonable to exclude those probes from the analyses (and is it possible)? TIA, Stefano