[BioC] Limma: All probes come out as significant

[Daniel Brewer]

Hello,

I have a curious problem involving Limma.  I have an ExpressionSet
object (called Seminoma) that contains the results of 18 samples (12
tumours and 6 normals).  The only strange thing I have done is to join
Affymetrix U133A and B results (renaming the probes so that there is no
overlap).

> design
 [1] 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0

I run the following to get the adjusted p-values for differential
expression.
> fit <- lmFit(Seminoma,design)
> fit <- eBayes(fit)
> tempo <- topTable(fit,n=nrow(exprs(Seminoma)),adjust="BH")
> tail(tempo)
             ID    logFC  AveExpr        t     P.Value   adj.P.Val         B
36857 236888_at 4.050545 5.896910 2.848388 0.007679017 0.007679872 -2.489166
39997 240028_at 3.590545 5.153103 2.795185 0.008761283 0.008762063 -2.602027
5638  206112_at 4.702957 7.202596 2.749623 0.009800000 0.009800654 -2.697717
37129 237160_at 3.356252 4.824457 2.721845 0.010488579 0.010489046 -2.755610
9742  210262_at 4.657111 7.232671 2.690246 0.011326586 0.011326838 -2.821047
23669 223673_at 3.837049 5.837806 2.648655 0.012524609 0.012524609 -2.906478

As you can see, all the probes appear to be significantly differentially
expressed.  I am sure this should not be the case, especially after
examining a number of different probes.  For example, on probe "117_at"
if I run a t-test() it produces a p-value of 0.1034 (no adjustment)
whereas limma suggests it is 8.4e-07 (or 2.45e-06).

My only thought is that something must be happening in the eBayes step.
 Can anyone help me on what I might be doing wrong?

Many thanks

Dan

PS
> sessionInfo()
R version 2.4.1 (2006-12-18)
x86_64-unknown-linux-gnu

locale:
LC_CTYPE=en_GB;LC_NUMERIC=C;LC_TIME=en_GB;LC_COLLATE=en_GB;LC_MONETARY=en_GB;LC_MESSAGES=en_GB;LC_PAPER=en_GB;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=en_GB;LC_IDENTIFICATION=C

attached base packages:
[1] "tools"     "stats"     "graphics"  "grDevices" "utils"     "datasets"
[7] "methods"   "base"

other attached packages:
 Biobase    limma
"1.12.2" "2.9.13"


-- 
**************************************************************
Daniel Brewer, Ph.D.
Institute of Cancer Research
Email: daniel.brewer at icr.ac.uk
**************************************************************

The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company Limited by Guarantee, Registered in England under Company No. 534147 with its Registered Office at 123 Old Brompton Road, London SW7 3RP.

This e-mail message is confidential and for use by the addre...{{dropped}}