Seurat: Do I have to run first RunUMAP or FindClusters?

[McStber]

Dear all, dear Seurat team,

I have two general questions.

In your vignettes "Integrating stimulated vs. control PBMC" your are running first RunUMAP and then FindNeighbors (setting the reduction to "pca" so that it will not take the UMAP reduction).
1) For running "FindClusters" you do not specify "reduction = pca" because the clustering is already done with "pca"??? Do I have to add reduction = pca here as well?

immune.combined <- RunUMAP(immune.combined, reduction = "pca", dims = 1:20)
immune.combined <- FindNeighbors(immune.combined, reduction = "pca", dims = 1:20)
immune.combined <- FindClusters(immune.combined, resolution = 0.5)

In the Vignette "Guided Clustering Tutorial" you are running RunUMAP after FindingClusters:
pbmc <- FindNeighbors(pbmc, dims = 1:10)
pbmc <- FindClusters(pbmc, resolution = 0.5)
pbmc <- RunUMAP(pbmc, dims = 1:10)

2) Is that because you are using UMAP here only for visualization? Why is the order different here compared to the other Vignitte?
Just want to understand it =)

Best regards,
Michael

[andrewwbutler]

Hi Michael,

1. FindClusters performs graph-based clustering on the neighbor graph that is constructed with the FindNeighbors function call. This neighbor graph is constructed using PCA space when you specifiy reduction = "pca". You shouldn't add reduction = "pca" to FindClusters.

2. Yes, UMAP is used here only for visualization so the order of RunUMAP vs FindClusters doesn't really matter (you just want to have run both before plotting with DimPlot in order to visualize the clusters in UMAP space).

[singletrips]

If I try different resolutions with FindClusters(), should I run FindNeighbors() multiple times or I only need to run FindNeighbors() only once?

[andrewwbutler]

If you are only changing the resolution parameter, you only need to rerun FindClusters.