[BioC] question about limma2annaffy

[James W. MacDonald]

Hi Jenny,

Jenny Drnevich wrote:
> Hi Jim,
> 
> I'm learning to use your limma2annaffy function, and I have a few 
> questions. In the affycoretools vignette (Feb 6, 2007), you call 
> limma2annaffy this way:
> 
>  > limma2annaffy(eset, fit2, design, annotation(eset), pfilt = 0.05)
> 
> I think there's a typo in this, because you've left out the contrast 
> matrix. When I try the same on my data:

Yeah, that's a typo. It never came up in a build/check cycle because it 
never really gets called in the Sweave() document. Thanks for pointing 
it out though.

> 
>  > limma2annaffy(gcrma.pres, fit2, design, annotation(gcrma.pres),pfilt=0.05)
> Error in vector("list", dim(contrast)[2]) :
>          negative length vectors are not allowed
> 
> When I try naming the annotation(gcrma.pres) argument, I get this:
> 
>  > limma2annaffy(gcrma.pres, fit2, design, 
> lib=annotation(gcrma.pres),pfilt=0.05)
> Error in vector("list", dim(contrast)[2]) :
>          argument "contrast" is missing, with no default
> 
> It does work when I add the contrast matrix:
> 
>  > limma2annaffy(gcrma.pres, fit2, design, 
> cont.matrix,annotation(gcrma.pres),pfilt=0.05)
> 
> 
> So, besides pointing out an apparent error in the vignette, I had a 
> question as to what you would do if you didn't use a contrast matrix? Now, 
> most all the time that I analyze affy data I do use a contrast matrix, but 
> you don't always have to use one. I have some other questions/suggestions 
> to improve the ease of use of affycoretools, but they are too complicated 
> to put in this e-mail...

You can _always_ use a contrasts matrix, even if you don't need one. The 
thing about limma2annaffy() is I use the colnames of the contrasts 
matrix to name the output files, so there has to be one. So there are 
two things you can do.

Say you fit a factor effects model with a set of tumor/normal samples, 
three of each:

 > design <- model.matrix(~ factor(rep(1:2, each=3)))
 > design
   (Intercept) factor(rep(1:2, each = 3))2
1           1                           0
2           1                           0
3           1                           0
4           1                           1
5           1                           1
6           1                           1
attr(,"assign")
[1] 0 1
attr(,"contrasts")
attr(,"contrasts")$`factor(rep(1:2, each = 3))`
[1] "contr.treatment"

Now the second column of the design matrix specifies the tumor - normal 
contrast, so a contrasts matrix is superfluous. However, this contrast 
will work with limma2annaffy():

 > contrast <- matrix(c(0,1), dimnames=list(c("normal", "tumor"), "tumor 
vs normal"))
 > contrast
        tumor vs normal
normal               0
tumor                1


Or you can just specify a cell means model in the first place.

Alternatively, you could just select the probesets that are significant 
and use probes2table(). But that would be more work if you want t-stats, 
p-values, etc.


If you want to send me an email offline, I would be happy to hear your 
suggestions. I am always up for improving things.

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.