PhILR to transformAssay

[TuomasBorman]

Should we add PhilR to transformAssay so that user can apply the transformation similarly to others?

transformAssay(tse, method = "philr")

Currently, it needs to be done "manually"

philr::philr(assay(tse, "counts"), rowTree(tse))

[antagomir]

The problem is that this transformation changes data dimensions and this does not with assays.

[TuomasBorman]

Does it make sense to

a) Give warning and subset TreeSE to match the new dimensions
b) Create a new TreeSE with these new dimensions, and output it inside altExp (probably too complex)
c) Force the new assay to the original dimensions, and fill the values with NA

[antagomir]

The new features do not match with the original features (= a requirement of an assay). The phILR features correspond to ratios between taxa pairs and this is fundamentally different from abundances of individual taxa.

Forcing assay to accommodate this does not sound a valid approach to me.

Therefore it should be something like altExp where this is stored if a solution is to be found. But then it is not so clear that this is outcome from transformation. Perhaps that's ok?

This question is the reason why we do not have a clear way to incorporate philr yet, there has been no consensus how to do this. But it is a good question, I wish we could come up with a good solution.

[TuomasBorman]

I haven't familiarized myself with the method, sorry about that...

I think it might make sense to

1. Add philr transformed data to altExp
2. Give message: "philr data is in altExp".

[antagomir]

Works for me. It means that philr pkg would become a new dependency.

The manpage entry could link to philr package and its manuals for more examples.

Leo