[Bioc-devel] why doesn't this ``just work'' ? (defined coercion is ignored by S3 method?)

[Hervé Pagès]

Hi Tim,

On 02/26/2013 02:01 PM, Tim Triche, Jr. wrote:
> It seems like the best way to write an email around these parts is using R
> code, so here goes.  I'm just wondering why a coercion that I define for
> DataFrame -> data.frame doesn't automatically get used in practice.
>
>
> library(Biobase)
> library(GEOquery)
> library(GenomicRanges)
>
> ## download a dataset which will become a SummarizedExperiment
> ##
> gset <- getGEO("GSE41826") ## GSEMatrix seems to hose it
> if (length(gset) > 1) idx <- grep("GPL13534", attr(gset, "names"))
>    else idx <- 1  ## a kludge, for dealing with SuperSeries entries
> gset <- gset[[idx]]
> sampleNames(gset) <- gset$title
>
> ## turn it into a SummarizedExperiment so I can call DMRs
> if(require(regulatoR))
>    sortedBrainCells <- as(gset, 'SummarizedExperiment')
>    ## the above could maybe become a generic for GEOquery... !?
>
> ## massage some covariates
> sortedBrainCells$sex <-
>      as.factor(sub('Sex: ', '', sortedBrainCells$characteristics_ch1.2))
> sortedBrainCells$race <-
>      as.factor(sub('race: ', '', sortedBrainCells$characteristics_ch1.3))
> sortedBrainCells$age <-
>      as.numeric(sub('age: ', '', sortedBrainCells$characteristics_ch1.4))
>
> ## now the part that is bugging me: why won't R coerce automatically?
> ##
> ## define a coercion for DataFrame to data.frame:
> setAs("DataFrame", "data.frame", function(from) as.data.frame(from))
> ##
> ## try to assemble a model.matrix
> model.matrix(~ age + race + sex, data=colData(sortedBrainCells))
> ##
> ## D'OH!
> ##
> ## Error in as.data.frame.default(data, optional = TRUE) :
> ##  cannot coerce class "structure("DataFrame", package = "IRanges")" to a
> data.frame

What seems to be called internally is as.data.frame.default, and
of course as.data.frame.default as no idea how to convert a DataFrame
into a data.frame:

   > as.data.frame.default(DataFrame())
   Error in as.data.frame.default(DataFrame()) :
     cannot coerce class "structure("DataFrame", package = "IRanges")" 
to a data.frame

I think this would work if you defined an S3 as.data.frame method
for DataFrame, instead of an S4 "coerce" method from DataFrame to
data.frame. Try to define the following:

   as.data.frame.DataFrame <- selectMethod("as.data.frame", "DataFrame")

A couple of months ago I've started to add some S3 methods to the
IRanges/GenomicRanges infrastructure. So far I only did it for
duplicated, unique, sort, levels, as.list, window, window<-. but
many are still missing (30-40 or more). as.data.frame.DataFrame is
one of them.

Cheers,
H.

>
> ##  ok fine whatever, I'll do the coercion myself
> mat <- model.matrix(~ age + race + sex,
> data=as(colData(sortedBrainCells),'data.frame'))
>
> head(mat)
> ##        (Intercept) age raceAsian raceCaucasian sexMale
> ## 5175-G           1  47         0             1       1
> ## 5175-N           1  47         0             1       1
> ## 813-N            1  30         0             1       0
> ## 1740-N           1  13         0             0       0
> ## 1546-N           1  14         0             0       0
> ## 1546-G           1  14         0             0       0
>
>

-- 
Hervé Pagès

Program in Computational Biology
Division of Public Health Sciences
Fred Hutchinson Cancer Research Center
1100 Fairview Ave. N, M1-B514
P.O. Box 19024
Seattle, WA 98109-1024

E-mail: hpages at fhcrc.org
Phone:  (206) 667-5791
Fax:    (206) 667-1319