Skip to content

[Bioc-devel] [question about class extension]

4 messages · Delphine Charif, Sean Maden, migdal migdal

Delphine Charif
# 
Dear Bioc developer,

We are developing a package and have used the MultiAssayExperiment class to manage our data.
We need to impose a particular structure on the slot metadata of the MAE object (list => named list)
because we havec defined methods that depend on this named list.
Is it good practice to extend the MAE class for the specificity of our metadata structure or is it better
to define new methods applying to a MAE object  (and check the metadata structure )?

Thanks in advance,






--------------------------------------
Delphine CHARIF
+33 (0)1 30 83 35 24

IJPB<https://ijpb.versailles.inrae.fr/> - <https://www.inrae.fr/> INRAE<https://www.inrae.fr/>
RD10 - Route de Saint-Cyr
78026 Versailles
---------------------------------------
En temps partiel le mercredi
Sean Maden
# 
There aren't strict standards imposed on many of the packages in
Bioconductor introducing classes, but many other packages extend and impose
standards and constraints for given areas/fields/subjects.

If you are considering this route, something like the following could work
to determine if introducing metadata features and restrictions is the route
to take:

libraryVector <-
  c("MultiAssayExperiment", "SummarizedExperiment",
    "SingleCellExperiment", "SpatialExperiment")

# Try something like this to test your candidate data classes:

testClass <- function(libraryName, functionString){
  # libraryName : valid class library
  eval(parse(text = paste0(functionString,"(",libraryName,")")))
}

for(libraryName in libraryVector){
  testClass(libraryName, "library")
  testClass(libraryName, "new <- ")

  metadata(new) #

  testClass("new", "rm")
  testClass(libraryName, "detach")
}

There may be better ways to harmonize and share the data with collaborators
as a flat table.

MAE has extensive documentation and there is also a cheatsheet provided to
help out with tasking.

See also: DOI: 10.18129/B9.bioc.MultiAssayExperiment

Sean

On Tue, Jan 23, 2024 at 5:45?AM Delphine Charif <delphine.charif at inrae.fr>
wrote:

  
  


  


      

    

    

      
      

        


  

        

      migdal migdal
    

    

      
      #
    

  


  

      
Hi Delphine,

I had a similar dilemma while developing midasHLA
<https://github.com/Genentech/midasHLA> package. There, we wanted to have
unique features names across all experiments and colData as well as
mandatory metadata. To this end I wrote a new class specific to my package
that extends MAE. I had some negative comments
<https://github.com/Genentech/midasHLA/issues/5#issue-727826240> regarding
this solution (see R code section of the comment by LTLA; it might help you
decide how to approach your particular case), but IMO it works nicely.

Best Maciek

        
On Wed, 24 Jan 2024 at 14:43, Sean Maden <maden.sean at gmail.com> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Sean Maden
    

    

      
      #
    

  


  

      
Pleased to try also:

```
sapply(c("SummarizedExperiment", "MultiAssayExperiment"), library,
character.only = T)
new <- SummarizedExperiment()
metadata(new) <- list(MultiAssayExperiment())
new
```

Again, it is a preference.

kind regards,

Sean

        
On Wed, Jan 24, 2024 at 6:06?AM migdal migdal <mcjmigdal at gmail.com> wrote: