Strong impact of number of ADT PCs on WNN UMAP?

[erflynn]

Reposting from Issues #9482

I'm running WNN following the tutorial on a dataset of about 200,000 skin cells. I noticed that choice of number of ADT PCs has a huge impact on the UMAP -- the UMAP clearly looks better with more ADT PCs, despite the elbow plot indicating we should likely use fewer. I am using the same exact settings (30 RNA PCs) except the number of ADT PCs.

Using the bmcite dataset and code in the (WNN tutorial)[https://satijalab.org/seurat/articles/weighted_nearest_neighbor_analysis], we also see the pattern that too few PCs from ADT or RNA lead to poor UMAPs; however, the numbers at which this occurs are much lower (e.g. 3 or 5 PCs rather than 10 or 15). It does seem to be the case that after "enough" PCs, it looks relatively similar? (Though in other cases, particularly in sub-clustering, we've found that "too many" PCs make the UMAP look worse.)

Has anyone else found this? This kind of goes opposite of my initial intuition -- if WNN is robust to noise in the data, then if we give it "less" ADT info, it should look better but not worse? My hypothesis is this is a dimensionality issue -- at too low numbers of PCs, in ADT space, cells look artificially "closer", making it seem like the ADT is more helpful? I've also noticed that at higher numbers of ADT dimensions, the UMAP just starts looking exactly like the RNA UMAP.