Hi, I have a problem with an exprSet, that consists of 16 samples with 2 replicates each, i.e. 32 arrays. Using > exprs(eset) I get the expression values for each gene in each array, with the two replicates as adjacent columns, like this: A1 A2 B1 B2 C1 C2 D1 D2 ... I would like to calculate the mean of the two replicates for each gene and generare a matrix of the mean values. How can I do this? Likewise, using >Calls<-mas5calls(AffyBatch) >exprs(Calls) I get a data frame with mas5 calls (P, A, or M). I would like to test, if the calls for the two replicates are the same and return the call to a data frame, otherwise return NA. Any idea how to do this? Your help will be highly appreciated! Georg

On 11/29/05 12:26 PM, "Georg Otto" <georg.otto at="" tuebingen.mpg.de=""> wrote: > Hi, > > I have a problem with an exprSet, that consists of 16 samples with 2 > replicates each, i.e. 32 arrays. > > Using >> exprs(eset) > > I get the expression values for each gene in each array, with the two > replicates as adjacent columns, like this: > > A1 A2 B1 B2 C1 C2 D1 D2 ... > > I would like to calculate the mean of the two replicates for each gene > and generare a matrix of the mean values. How can I do this? Hi, Georg. Instead of averaging, I would suggest using a method of analysis that allows you to appropriately replicates as such. Look at limma and using the block argument. > >> Calls<-mas5calls(AffyBatch) >> exprs(Calls) > > I get a data frame with mas5 calls (P, A, or M). I would like to test, > if the calls for the two replicates are the same and return the call > to a data frame, otherwise return NA. Any idea how to do this? Again, I would try to use all the data as best you can. You could set values in your expression matrix to NA or downweight probesets that have an absent call if you are using limma. There are many ways to do these things, but I think averaging and other "lumping" techniques may not be the right way to go. Sean

Dear Sean, thanks a lot for your advice! In principle I completely agree, and I often use limma for blocking, downweighting etc. However, somtimes I would like to do some quick filtering of genes, eg. to find out how many are above a certain intensity level, how many have present calls in more than x experiments etc. Best, Georg Sean Davis <sdavis2 at="" mail.nih.gov=""> writes: > On 11/29/05 12:26 PM, "Georg Otto" <georg.otto at="" tuebingen.mpg.de=""> wrote: > >> Hi, >> >> I have a problem with an exprSet, that consists of 16 samples with 2 >> replicates each, i.e. 32 arrays. >> >> Using >>> exprs(eset) >> >> I get the expression values for each gene in each array, with the two >> replicates as adjacent columns, like this: >> >> A1 A2 B1 B2 C1 C2 D1 D2 ... >> >> I would like to calculate the mean of the two replicates for each gene >> and generare a matrix of the mean values. How can I do this? > > Hi, Georg. > > Instead of averaging, I would suggest using a method of analysis that allows > you to appropriately replicates as such. Look at limma and using the block > argument. > > >> >>> Calls<-mas5calls(AffyBatch) >>> exprs(Calls) >> >> I get a data frame with mas5 calls (P, A, or M). I would like to test, >> if the calls for the two replicates are the same and return the call >> to a data frame, otherwise return NA. Any idea how to do this? > > Again, I would try to use all the data as best you can. You could set > values in your expression matrix to NA or downweight probesets that have an > absent call if you are using limma. > > There are many ways to do these things, but I think averaging and other > "lumping" techniques may not be the right way to go. > > Sean