Sun, Yezhou wrote: > Hello, Jim, > > Thank you very much for all the suggestions. The problem was solved. > However I got several new questions: > > 1. what is the exact meaning of -1, 0, 1 in design matrix? It would take more than an email to answer this question. However, you are in luck because Natalie Thorne has posted some really nice slides that should help you to figure this out. http://www.damtp.cam.ac.uk/user/npt22 > > 2. I also ran SAM on this data set. There is almost no overlapping > between top genes picked up by SAM and limma. What could this imply? Most likely that you have fit a different model in SAM. In my experience you don't get wildly different results from SAM and limma, but it might not be obvious how to do the same analysis using the two different tools. > > 3. It seems that lmFit requires the number of columns in data set be > same as the number of arrays in design matrix. So there is no way to put > probe ID into data set. There are numbers listed in ProbeID field in > output from topTable(). Are these numbers supposed to be row ID in data > set? You *can* put the probe IDs into the data set by setting the row.names() of your matrix to be the probe IDs. As an alternative, you can use the genelist argument to topTable(). See ?topTable for more information. > > My data set is from GE Codelink platform and has been median centered > and log2 transformed. Median centered? You can do much more sophisticated normalizations using limma than a simple median centering. Are you sure that this is all you need? > > Thanks a lot. > > Yezhou Jim -- James W. MacDonald University of Michigan Affymetrix and cDNA Microarray Core 1500 E Medical Center Drive Ann Arbor MI 48109 734-647-5623 ********************************************************** Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues.

Hello, Jim, Thank you very much for all the suggestions. The problem was solved. However I got several new questions: 1. what is the exact meaning of -1, 0, 1 in design matrix? 2. I also ran SAM on this data set. There is almost no overlapping between top genes picked up by SAM and limma. What could this imply? 3. It seems that lmFit requires the number of columns in data set be same as the number of arrays in design matrix. So there is no way to put probe ID into data set. There are numbers listed in ProbeID field in output from topTable(). Are these numbers supposed to be row ID in data set? My data set is from GE Codelink platform and has been median centered and log2 transformed. Thanks a lot. Yezhou -----Original Message----- From: James W. MacDonald [mailto:jmacdon@med.umich.edu] Sent: Friday, February 24, 2006 5:31 PM To: Sun, Yezhou Cc: BioConductor_list; Zhang, Weijia Subject: Re: [BioC] limma question Sun, Yezhou wrote: > Hi, Jim, > > I tried what you suggested but it's still not working. Got same error > message: > > >>fit<-lmFit(d[, -1], design) > > Error in qr(x) : NA/NaN/Inf in foreign function call (arg 1) > In addition: Warning message: > NAs introduced by coercion > > > I think the problem is still in data frame but I don't know what it is. > The user guide has no detailed description for single channel data set. Affy data are single channel, and I think there are two or three examples in the user guide. lmFit() is expecting some sort of BioC type object (exprSet, MAlist, marrayNorm, PLMset) or a numeric matrix. A matrix can only contain data of one type (character, factor, numeric, etc), whereas a data.frame can contain data of mixed types. One of the first things lmFit() is doing to your data.frame is coercing to a matrix, which I think is the reason for the Warning message saying NAs introduced by coercion. I figured that the first column was the problem, but you need to see if there are any other NAs in the data.frame, or better yet do new.d <- as.matrix(d[,-1]) and look for the NAs. Best, Jim -- James W. MacDonald University of Michigan Affymetrix and cDNA Microarray Core 1500 E Medical Center Drive Ann Arbor MI 48109 734-647-5623 ********************************************************** Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues.