Hi, I hope someone can help me out. I am using Bioconductor and limma to examine my microarrat data. What I am trying to do is eliminate any spots that have an NA between arrays before applying the design matrix in limma and applying the eBayes function. Even after I have done this elimination process and have checked it, by having a look at the M values of the top genes in the topTable list, some of the genes still have an NA value for one array, but designated M values for the rest of the arrays in the group. This is what I am typing into the workspace after normalisation and scalenormalisation: A.object<-maA(objectscalenormalisation) M.object<-maM(objectscalenormalisation) Library(limma) a<-is.na(M.object) a<-apply(a,1,sum) b[b==0] which[b==0] zero<-which(b==0) x<- list(A=maA(objectscalenormalisation)[zero,],M=maM(objectscalenormalisa ti on)[zero,] x.lmFit<-lmFit(x,c(1,-1,1,-1)) x.lmFit.Bayes<-eBayes(x.lmFit) plot(x.lmFit.Bayes$coef,x.lmFit.Bayes$lods) topTable(x.lmFit.Bayes,n=100) On Wednesday, July 7, 2004, at 04:15 PM, James Wettenhall wrote: Why might this not be working ? Michael. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: text/enriched Size: 1195 bytes Desc: not available Url : https://www.stat.math.ethz.ch/pipermail/bioconductor/attachments /20040709/99a9ec52/attachment.bin

I believe the main problem is that you switch your variable name from a to b without carrying anything over. Maybe something like this would work? a <-is.na(M.object) zero <- apply(a,1,sum) == 0 Then continue on from there. HTH, Jim James W. MacDonald Affymetrix and cDNA Microarray Core University of Michigan Cancer Center 1500 E. Medical Center Drive 7410 CCGC Ann Arbor MI 48109 734-647-5623 >>> Michael Stapelberg <msberg@med.usyd.edu.au> 07/08/04 11:00PM >>> Hi, I hope someone can help me out. I am using Bioconductor and limma to examine my microarrat data. What I am trying to do is eliminate any spots that have an NA between arrays before applying the design matrix in limma and applying the eBayes function. Even after I have done this elimination process and have checked it, by having a look at the M values of the top genes in the topTable list, some of the genes still have an NA value for one array, but designated M values for the rest of the arrays in the group. This is what I am typing into the workspace after normalisation and scalenormalisation: A.object<-maA(objectscalenormalisation) M.object<-maM(objectscalenormalisation) Library(limma) a<-is.na(M.object) a<-apply(a,1,sum) b[b==0] which[b==0] zero<-which(b==0) x<- list(A=maA(objectscalenormalisation)[zero,],M=maM(objectscalenormalisa ti on)[zero,] x.lmFit<-lmFit(x,c(1,-1,1,-1)) x.lmFit.Bayes<-eBayes(x.lmFit) plot(x.lmFit.Bayes$coef,x.lmFit.Bayes$lods) topTable(x.lmFit.Bayes,n=100) On Wednesday, July 7, 2004, at 04:15 PM, James Wettenhall wrote: Why might this not be working ? Michael.

Hi Michael, On Fri, 9 Jul 2004, Michael Stapelberg wrote: > A.object<-maA(objectscalenormalisation) > M.object<-maM(objectscalenormalisation) You really need to explain what you did from the beginning. The R commands you have presented begin with the functions maA() and maM(). If they are from the "marray" package then you should say that explicitly, i.e. you should say that you used: library(marray) > a<-is.na(M.object) If you are worried about background-corrected red or green intensities being negative and log ratios then being undefined or in other words Not Available (NA), then this "problem" is certainly not unique to your data set. The authors of the marray and limma packages are well aware of this issue and there are automatic methods to deal with this. You shouldn't need to manually check for NAs for every data set. I would suggest going through the examples in the limma and marray documentation, and trying to apply the steps to your data set as closely as possible. If you are using limma, you will create an MAList object (the usual name given to this object is "MA"). If you are using the marray package, you will instead create an marrayNorm object (which contains M and A objects after normalization, within an S4 object). You seem to be creating M and A using a non-standard method (or extracting them from an marrayNorm object unnecessarily), so this makes it difficult for people to help you. If you want to convert an marrayNorm object (from the marray package) into an MAList object (for the limma package), in order to fit a linear model, there are automatic coercion (conversion) functions to do this in the convert pacakge, i.e. you don't need to extract M and A manually. To convert an marrayNorm object to an MAList object, you can do: library(convert) mnorm <- new("marrayNorm") MA <- as(mnorm,"MAList") Replace "mnorm" above with your object of class marrayNorm. Regards, James