Actually automatically (using heuristics/etc.) selecting features for a given ranking in the visualization

[fedarko]

Based on conversation with @ElDeveloper.

Basically, this would involve solving some sort of optimization problem where you want to select "stable" features across samples, and find features that are highly associated with a given ranking column. This is sort of akin to what selbal does, but just using the rankings (from DEICODE/Songbird/ALDEx2/...) as a guide instead of actually looking at the log-ratio's associations with a metadata category.

[fedarko]

@mortonjt, figured I should ask you if you have any thoughts on this sort of approach (have you tried something like this in the past).

Implementing this could be a nice goal for the next release of Qurro?