Seurat: add customized clustering info to seurat object

[crazyhottommy]

Hi,

With the scaled (regressed) seurat object, how can I add K-means clustering or hierarchical clustering info to it?

# Create and setup Seurat objects for each dataset
WM115 <- CreateSeuratObject(raw.data = WM115.data, project = "WM115")
WM115 <- FilterCells(WM115, subset.names = c("nGene", "percent.mito"), 
    low.thresholds = c(500, -Inf), high.thresholds = c(Inf, 0.15))

WM115 <- NormalizeData(WM115)
WM115 <- FindVariableGenes(WM115 do.plot = F, display.progress = F)

## always scale z score for PCA
WM115<- ScaleData(object =WM115, vars.to.regress = c("nUMI", "percent.mito"))

# graph based clustering
WM115 <- FindClusters(WM115,
    resolution = 0.6, dims.use = 1:10)

## how to add k-means and others?
....

Thanks!

Tommy

[crazyhottommy]

essentially I want to cluster single cells using k-means and color the points in the tsne plot. now the clustering is based on a graph-based (KNN) method, and do differentially expression based on the k-means clusters

[leonfodoulian]

Hi Tommy,

If you have already computed these clustering independently, and would like to add these data to the Seurat object, you can simply add the clustering results in any column in [email protected]. You can then set the clustering results as identity of your cells by using the Seurat::SetAllIdent() function. For an example on how to use this function, you can check issue #325. This should allow your cells to be coloured accordingly when plotting the tSNE embeddings using Seurat::TSNEPlot(). An alternative is to pass the column name storing these data in [email protected] as input to the group.by argument of Seurat::TSNEPlot().

# Add k-means clustering results to '[email protected]$k.means.clusters'
[email protected]$k.means.clusters <- k.means.result

# Plot tSNE, and colour cells by k-means identity
TSNEPlot(object = object, group.by = "k.means.clusters")

Best,
Leon

[crazyhottommy]

Thank you! this helped!

[thomasveith]

I am trying to do something similar

# Add our clonal groupings
[email protected]$membership <- membership

# Set cell identity to sample identity
katoiii_s <- SetIdent(object = katoiii_s, value = "membership")

# tSNE plot
katoiii_s <- TSNEPlot(object = katoiii_s, group.by = "membership")

And I get the error:

Error in `[[.Seurat`(object, reduction) : 
  Cannot find 'tsne' in this Seurat object

[ibseq]

Hi
can we import customised clustering info from 10x (csv file with each barcode assigned to a cell type = cluster) into seurat, if yes, how can this be done?

thanks
ib

[bounlu]

@crazyhottommy

You can also do like:

object$kmeans_10 = kmeans(x = object@reductions[["pca"]]@cell.embeddings, centers = 10, iter.max = 100, nstart = 100)$cluster

DimPlot(object, reduction = "tsne", label = T, group.by = "kmeans_10", split.by = "orig.ident") + NoLegend()

Idents(object) = "kmeans_10"

FindAllMarkers(object, assay = "RNA", only.pos = TRUE)