[Bioc-devel] eSet class and two color arrays

Robert Gentleman rgentlem at fhcrc.org
Fri Jun 2 19:01:37 CEST 2006


Wolfgang Huber wrote:
> Hi Martin, Seth, & others,
> one of the basic features of the eSet class is that
> - it has exactly one phenoData slot, and >=1 assayData slots
> - the rows of the dataframe in phenoData correspond 1:1 to the columns
>   of each the assayData matrices.
> I was wondering whether there is interest in a class that has
> - one or several phenoData slots, and the same number of assayData slots
> - there is a 1:1 correspondence between these
> - the rows of each phenoData correspond to the columns
>   of the corresponding assayData matrix.

  Why not a list of eSets? Why do you want to put these together in this 

> Andreas Buness from arrayMagic has a use case where such a class could
> be used to represent raw two-color microarray data.
> Alternatively, it is also possible to put two-color data into the
> existing eSet design, if a convention is used that each array is
> represented by two rows in phenoData (corresponding to the two colors)
> and two columns in assayData. This is essentially what now happens in
> the exprSetRG class of arrayMagic. So a useful eSet-based class for
> two-color data (formalizing the convention) could still be derived from
> the existing eSet definition. But I wanted to put up this for
> discussion: do other people have a need/wish for "multiple phenoData"
> design?

   We decided not to deal with preprocessed data (I think). And that 
eSet corresponds to expression, not raw data. So when you get here, you 
do not have red or green channels, you have expression (relative?) not 
raw data. Packages like limma, marray etc, I believe have their own 
internal formats for the raw (two channel data) and the goal of the eSet 
classes in Biobase is to provide a uniform data structure for downstream 
analysis. While one can see obvious similarities; one channel data also 
has different, but similar, structures for preprocessed data than 
processed data -- there too we decided not to get involved at the 
preprocessing stage -- eSet and friends are for post-processed data so 
that it can be treated uniformly by tools for differential expression, 
machine learning etc.

  If a special preprocessed data class is wanted, then I think it would 
be better to define it on its own and not try to leverage the eSet class 
into double duty.

  Now, you might want to use, phenoData, AssayData and other building 
blocks in Biobase, to construct the classes you describe, just not eSet, 
it has a role in life.

   best wishes

> Cheers
>  Wolfgang

Robert Gentleman, PhD
Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M2-B876
PO Box 19024
Seattle, Washington 98109-1024
rgentlem at fhcrc.org

More information about the Bioc-devel mailing list