[BioC] limma and ebayes : error message : differences in versions?

[James W. MacDonald]

Hi Sue,

Sue Jones wrote:
> 
> 
> I have used limma to calculate differential gene expression (see code at 
> end of message). I am currently using R 2.7, Biobase 2.0.1 and Limma 
> 2.14.6 and get the error message
> 
> Warning message:
> Zero sample variances detected, have been offset

That's not an error - it's a warning. The difference being that your 
analysis proceeded, but there was something you might want to know about 
versus your computer going crash bang thud and halting.

> 
> after uisng eBayes.
> 
> I know a previous message to this list explains this and says it can be 
> ignored.
> 
> However, I conducted the same analysis using older versions of R, Limma 
> and Biobase and did not see this message and got a much larger list of 
> differentially expressed genes. Using the current versions of limma and 
> biobase I see this message and my list of differentially expressed genes
> is smaller - can this be correct?

Yes, but probably for a reason different from the one you mention. You 
are using gcrma() to generate expression values and there have been 
several comments on this list about more recent versions of gcrma() 
generating summary statistics that are identical for all samples (hence 
zero sample variance).

So, is this a problem? I don't know. It is possible that your previous 
list of differentially expressed genes had more false positives than 
your current list. Or not. That's the problem with much of this high 
throughput data - you get tons of data, but there is no ground truth so 
it is difficult to separate the signal from the noise.

Best,

Jim


> 
> Thanks for anyones help.
> 
> 
> Code
> ------------------------------------------------------
>> files <- c("W05h_1.CEL", "W0h_1.CEL", "W2h_1.CEL",
> + "W2h_2.CEL", "W4h_1.CEL",  "GW_025.CEL", "GW_1.CEL", 
> "GW_3.CEL","E2h_1.CEL",  "E2h_2.CEL",
> + "E2h_3.CEL", "E2h_4.CEL")
>> Data <- ReadAffy(filenames = files)
>> Data_gcrma <- gcrma(Data)
>> design<- cbind(WAT=c(1,1,1,1,1,1,1,1,0,0,0,0), 
>> ETH=c(0,0,0,0,0,0,0,0,1,1,1,1))
>> fit <- lmFit(Data_gcrma,design)
>> cont.matrix <- makeContrasts(WATvsETH=ETH-WAT, levels=design)
>> fit2<- contrasts.fit(fit,cont.matrix)
>> fit3 <- eBayes(fit2)
> Warning message:
> Zero sample variances detected, have been offset
> 
> 
> Sue Jones
> University of Sussex
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: 
> http://news.gmane.org/gmane.science.biology.informatics.conductor

-- 
James W. MacDonald, M.S.
Biostatistician
Douglas Lab
University of Michigan
Department of Human Genetics
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618
734-615-7826