View distal/promoter peaks after RNA/ATAC integration

[matthew-levy]

Is there a way to get the UMAP with distal vs promoter peaks on the integrated dataset? Similar to the function ac.pl.umap(atac, color=["KLF4"], average="peak_type") but using mu.pl.umap(mdata, color=["KLF4", "chr9:107480158-107492721"]).

[matthew-levy]

I guess a work around would to trick the ATAC AnnData object by copying the .obsm['X_umap'] from MuData after running MOFA. That way we could run the ac.pl.umap function but with the integrated UMAP coordinates. If I go this way, is there a way to store multiple sets of UMAP coordinates in an object and tell the plotting function which to use, or would I just need to overwrite it?

[gtca]

Thanks, @matthew-levy! Peak aggregation is probably going to stay a modality-specific feature: there's no way to know if this is the gene "KLF4" from RNA or peaks that are linked to this gene from ATAC that the user wants to visualise.

That being said, we can think of ways to make it a more general feature by exposing the interfaces accordingly, e.g. to plot aggregate values using a list of features.

The solution you mention would be the most transparent way to go. One can store and plot many embeddings:

atac.obsm["X_mofa_umap"] = mdata.obsm["X_umap"].copy()
ac.pl.embeddings(atac, basis="X_mofa_umap", color=["KLF4"], average="peak_type")