[BioC] rownames in topTable and posterior probability

Gordon Smyth smyth at wehi.edu.au
Mon Feb 28 05:46:13 CET 2005

At 01:45 PM 28/02/2005, Adaikalavan Ramasamy wrote:
>Thank you Dr. Symth (and Dr. MacDonald).
>As you have correctly suspected, the problem is with the input
>'onebatch' which is a matrix. The matrix was actually extracted from the
>exprSet during an intermediate step, stored and used for subsequent
>analysis here.
>Is it a sensible suggestion for the lmFit() to set the rownames of the
>matrix as the gene names _if_ the class of object was matrix ? Might try
>later today to fudge to include a branch for is(object, "matrix") in the
>existing lmFit() code and post it. But is this worth the effort ?

That's a sensible suggestion which is easy to implement. I'll put it in the 
next version of limma.

>I presume there might be a few people who prefer to work with matrix of
>gene expression. E.g. those who only get the finalised dataset as the
>output of some public repositories.

In the case of public repositories, I doubt that the matrix will have 
rownames attached.

Rownames are not in general a satisfactory way to store probe IDs because 
of the embarassment that non-unique rownames can potentially cause. The 
exprSet class gets away with storing probe IDs this way because exprSet is 
mostly used for Affy and Affy chips don't repeat probes.


>  Or is it easier to convert a matrix
>to an exprSet and work from there ?
>Thank you.
>Regards, Adai

More information about the Bioconductor mailing list