[Bioc-devel] eSet class and two color arrays

[Martin Morgan]

I guess I need some more info on the use case. As it is, there can be
several assayData elements. Each sample can have a corresponding
description in phenoData. phenoData can in turn have several columns
of variables, including variables that are only relevant to one of the
assayData elements. Subsetting works across assayData and phenoData in
a consistent way.

This seems to be sub-optimal if assayData elements are identically
dimensioned, but refer to different samples (??) or different parts of
phenoData refer to different assayData elements (??) . Is the issue
that, in these cases having separate columns of phenoData, one for
each set of assayData samples, seems artificial (though subsetting,
etc., work)?

Martin

> library(Biobase)
> data(sample.eSet) # old-style eSet
> mSet <- as(sample.eSet,"MultiSet")
Warning messages:
1: creating assayData colnames from phenoData sampleNames in:
updateOldESet(sample.eSet, "MultiSet")
2: creating assayData featureNames from reporterNames in:
updateOldESet(sample.eSet, "MultiSet")
> mSet
Instance of MultiSet 

assayData
  Storage mode: list 
  featureNames: geno2, geno3, 3XSSC, ..., 26-M5, 26-M9 (500 total)
  Dimensions:
          R   G  Rb  Gb
Rows    500 500 500 500
Samples   4   4   4   4

phenoData
  sampleNames: a, b, c, d
  varLabels:
    SlideNumber: number
    FileName: name
    Cy3: genotype labelled Cy3
    Cy5: genotype labelled Cy5
    Date: date

Experiment data
  Experimenter name:  
  Laboratory:  
  Contact information:  
  Title:  
  URL:  
  No abstract available.

Annotation character(0)
> mSet[1:10,c(1,4)]
Instance of MultiSet 

assayData
  Storage mode: list 
  featureNames: geno2, geno3, 3XSSC, ..., 3XSSC, 3XSSC (10 total)
  Dimensions:
         R  G Rb Gb
Rows    10 10 10 10
Samples  2  2  2  2

phenoData
  sampleNames: a, d
  varLabels:
    SlideNumber: number
    FileName: name
    Cy3: genotype labelled Cy3
    Cy5: genotype labelled Cy5
    Date: date

Experiment data
  Experimenter name:  
  Laboratory:  
  Contact information:  
  Title:  
  URL:  
  No abstract available.

Annotation character(0)
> nSet <- mSet
> assayData(nSet) <- assayDataNew(R=assayData(mSet)[["R"]], Rb = assayData(mSet)[["Rb"]])
> phenoData(nSet) <- phenoData(nSet)[,c("SlideNumber", "Cy3")]
> nSet
Instance of MultiSet 

assayData
  Storage mode: lockedEnvironment 
  featureNames: geno2, geno3, 3XSSC, ..., 26-M5, 26-M9 (500 total)
  Dimensions:
         Rb   R
Rows    500 500
Samples   4   4

phenoData
  sampleNames: a, b, c, d
  varLabels:
    SlideNumber: number
    Cy3: genotype labelled Cy3

Experiment data
  Experimenter name:  
  Laboratory:  
  Contact information:  
  Title:  
  URL:  
  No abstract available.

Annotation character(0)
> exprs(mSet) # which element of assayData should be the 'exprs'?
Error in exprs(mSet) : no direct or inherited method for function 'exprs' for this call

Wolfgang Huber <huber at ebi.ac.uk> writes:

> Hi Seth, Sean
>
>> So the use case is that one wants to subset all phenoData and
>> assayData components simultaneously?
>> 
>> I want to understand what features are desired beyond those provided
>> by a list of ExpressionSet instances (at first glance, a list seems
>> quite close to what you are asking for).
>
> More than a list:
> - simultaneous subsetting
> - consistency of dimensions and their content
>
>
> Sean Davis wrote:
>> I think the point Wolfgang is trying to make (feel free to clarify,
>> Wolfgang) is that the current phenoData and assayData constructs are
>> "Array-centric" while biologic questions can be asked that are
>> "sample-centric".  In other words, one often wants to treat
>> the two channels > on a single array as "separate samples" in a sense
>> (of course linked to each  other by array).
>
> Yes. An easy way out is to switch completely from array-centric to
> sample-centric (that's e.g. what arrayMagic does), but really one often
> wants to have both.
>
> Of course this whole discussion is not about what is possible to do
> (everything is possible with R) but how convenient it is.
>
>  Cheers W.
>
>
>
> -- 
> ------------------------------------------------------------------
> Wolfgang Huber  EBI/EMBL  Cambridge UK  http://www.ebi.ac.uk/huber
>
> _______________________________________________
> Bioc-devel at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/bioc-devel