Open artur-sannikov opened 1 year ago
If MOFA authors would welcome adding support for MAE this way then that would probably be the ideal solution.
Is there any limitation on how many experiments can be fed to MOFA? I am wondering that if there is no limitation, then there could be for instance 10 different experiments, which makes this altexp1, altexp2, ..., altexpn not scalable
I don't think there are limitations from technical point of view. In one of MOFA vignettes they used 5 experiments.
That makes the implementation of this more difficult. Although one option could be altexp parameter that takes a vector as input.
One additional annoying thing is that you cannot specify which experiments and assays you feed to MOFA. So you have to remove additional assays before running MOFA.
That could be also be fixed with vector parameter.
create_mofa_from_MultiAssayExperiment(
mae,
exps = c(1, 3)
assays = c("counts", "clr")
altexps = c(NA, 1)
)
Hi! In principle everything was implemented here: https://github.com/bioFAM/MOFA2/pull/144 Just waiting for a response for the past half year
Wow! You already did that :D
Hopefully the review proceeds soon
I will bump them again. This is a pity as the approved the idea originally, before the PR.
Is your feature request related to a problem? Please describe. Currently, to use
create_mofa_from_MultiAssayExperiment()
on agglomerated data, we have to directly modify the original MAE assays. In contrastgetExperimentCrossAssociation()
is capable of accepting alternative experiments (see here). I suggest implementing similar functionality forcreate_mofa_from_MultiAssayExperiment()
, which will provide us with more flexibility with running MOFA at different agglomeration levels.Describe the solution you'd like Add
altexp1
andaltexp2
parameters in a similar fashion togetExperimentCrossAssociation()
. @antagomir suggested to open a PR on he MOFA repo to add these parameters because this function belongs to MOFA2 package.Describe alternatives you've considered Implement this functionality in
mia
does not seem to be the most optimal option. The function belongs to MOFA and is more widely used thanmia
at the moment.