I know that Bluster leverages igraph to do clustering. Seurat has ordinary Louvain and "Louvain with multilevel refinement" (but no reference). Is there any way to get the second kind via Bluster? It does look nicer, as my supervisor noticed when my scater-based plots looked messier than a colleague who uses Seurat's plots, ignoring that it is all a specious art for just a wee moment.
I don't know what they're doing, but I'm happy to take a PR provided that the variant is simple to implement, i.e., builds on top of the existing igraph results without any further package dependencies.
If it's any more complicated, I really don't want to take on the extra maintenance burden.
FWIW I am skeptical about whether "plots looking nicer" should be a good metric for algorithmic decisions. But, if their variation is a genuine improvement over standard Louvain, it should probably be upstreamed into igraph's C library, rather than being trapped in the R ecosystem.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
If the plots 'look nicer' then this is probably due to UMAP parameters, and not the colour clustering as the clustering only comors indovidual points. Default UMAP parameters are a bit ugly indeed, try setting the spread and min_dist of UMAP to something like 0.75 to have points more spread out and not bunched up that much.