[BioC] affy troubles

[peter robinson]

=> Ken,
thanks. My main difficulty, which presumably comes from lack of sufficient 
experience with bioconductor, is that the object ex appeared to be different 
from the matrix object "golub", which is a matrix instance used in the 
vignette for the multtest package. This instance also has the attributes 
golub.gnames and golub.cl (a vector of tumor class labels). Also, the golub 
does not appear to have rownames as an attribute (presumably because it has 
.gnames).

So, I had assumed that exprs was supposed to return a comparable object. 

In the meantime, I have managed to get my analyses done (btw, thanks to all 
who have contributed to bioconductor, it is a truly amazing resource!), but I 
wound up creating a separate vector for the class labels and probably have 
done many things "correctly but clumsily".

So my question is whether there are settings to create matrix objects with a 
structure comparable to golub, and if this is the prefered way of working 
with bioconductor?

-Peter



Peter,

Your object "ex" is just a numeric matrix.  (to see this, do: class(ex)
and mode(ex).)  The row names of your matrix are probe set identifiers
(try e.g. rownames(ex)[1:10]).  NB: this is not the first column of the
matrix.  When you say "several of the downstream analyses do not work
properly", you should tell us which ones, and what errors you got.

Cheers,

Ken

On Mon, 17 Nov 2003, peter robinson wrote:

> Dear Bioconductors,
> 
> I would like to use the affy and multtest packages to extract a list of 
> differentially expressed genes from a set of 8 experiments, in which 4 are 
> wildtype and 4 are knockouts. I have worked through several of the 
affymetric 
> vignettes and the mult test vignette, which are all very well done. However, 
> for some reason, I cannot extract a proper n x m matrix with n= probe sets 
> and m = samples.
> My steps:
> 
> library(affy)
> Data<-ReadAffy()   /* Loads 8 .cel files with no error message  */
> eset <- rma(Data   /* also OK   */
> ex <- exprs(eset)  
> 
> I would like to get a matrix object similar to "golub" from the Vignette. 
> However, there are several differences. Firstly, "ex" has labels for the 
> probe_ids for all genes and does not display as a 12488 x 8 matrix (the 
> output of dim(ex), but rather as a (12488x8) x 2 matrix in which the first 
> column is the probe id and the second column is the expression level. Also, 
> there is no separate attribute such as golub.gnames.
> 
> Several of the downstream analyses as described in the multtest vignette do 
> not work properly, but I assume that the main problem is here.
> 
> Any ideas?
> 
> Thanks
> 
> Peter