How to use FindSubCluster in Seurat?

Question

[Pratik]

I just found the FindSubCluster tool within Seurat, and am super excited to use it. Just not sure exactly how!

The usage is here:

FindSubCluster(
  object,
  cluster,
  graph.name,
  subcluster.name = "sub.cluster",
  resolution = 0.5,
  algorithm = 1
)

I receive an error when using it.

> test <- FindSubCluster(scfp, 7, scfp, subcluster.name = "test",  resolution = 3, algorithm = 1)
Error in match(x, table, nomatch = 0L) : 
  'match' requires vector arguments
> test <- FindSubCluster(scfp, 7, scfjjd, subcluster.name = "test",  resolution = 3, algorithm = 1)
Error in `[[.Seurat`(object, graph.name) : object 'scfjjd' not found
> test <- FindSubCluster(scfp, 7, RNA, subcluster.name = "test",  resolution = 3, algorithm = 1)
Error in `[[.Seurat`(object, graph.name) : object 'RNA' not found

My object name is scfp. I have done FindClusters already. The cluster I wish to subcluster is 7.

I think my trouble is with the graph.name argument.

I tried this:

> test <- FindSubCluster(scfp, "7", "scfp", subcluster.name = "test",  resolution = 3, algorithm = 1)
Error: Cannot find 'scfp' in this Seurat object

Still no success yet...

Any help would really be appreciated!

[fracarb8]

I have never used that function, so I could be completely off, but have a look at FindMultiModalNeighbors. It could be similar to AddModuleScore, which computes the score and add a column (score name) to your metadata.

[Pratik]

Thank you @fracarb8. So I got the command to go through after this:

I went to: https://satijalab.org/seurat/articles/multimodal_reference_mapping.html

and Ctrl+F: graph.name

Turns out graph.name is an argument in FindNeighbors. So I just use the FindNeighbors command choosing graph.name as some arbitrary name (test). and then reran FindSubCluster and it worked, partially.

So it looks something like this:

scfp <- FindNeighbors(scfp, graph.name = "test", dims = 1:100)
scfp <- FindClusters(scfp, graph.name = "test", resolution = 4, algorithm = 1, verbose = TRUE)
scfp <- FindSubCluster(scfp, "5", "test", subcluster.name = "blood",  resolution = 2, algorithm = 1)
Modularity Optimizer version 1.3.0 by Ludo Waltman and Nees Jan van Eck

Number of nodes: 93
Number of edges: 758

Running Louvain algorithm...
0%   10   20   30   40   50   60   70   80   90   100%
[----|----|----|----|----|----|----|----|----|----|
**************************************************|
Maximum modularity in 10 random starts: 0.2673
Number of communities: 8
Elapsed time: 0 seconds
1 singletons identified. 7 final clusters.

However, no success in visualizing the results via tSNE or UMAP. I tried rerunning RunTSNE, and then seeing the plot, but still no success.

Answer 1

[Pratik]

Okay so I got it... I think. I explored the Seurat object a litle bit more and found that the cluster assignments were saved. I was able to visualize using the group.by argument so here is the process:

scfp <- FindNeighbors(scfp, graph.name = "test", dims = 1:100)
scfp <- FindClusters(scfp, graph.name = "test", resolution = 2, algorithm = 1, verbose = TRUE)
scfp <- RunTSNE(scfp, dims = 1:100)
DimPlot(scfp, reduction = "tsne", label = TRUE, label.size = 6 )

And then after doing this:

scfp <- FindSubCluster(scfp, "6", "test", subcluster.name = "blood",  resolution = 2, algorithm = 1)
DimPlot(scfp, reduction = "tsne", group.by = "blood", label = TRUE, label.size = 6)

This is the resulting tSNE:

Of course going to play around to optimize cluster numbers...

Thank you @fracarb8. You got me going!

[Pietro]

Just found out that even if you don't rerun FindNeighbors with graph.name = "test", you can find the default names of the graphs here: names(yourSeuratObject@graphs).

For my integrated object these are integrated_nn and integrated_snn.