This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Seurat re-clustering a cell subset but cell identity numbers are not completely showing up

Hi,

After running T-SNE plot on dataset, I tried to extract a specific cell type then do reclustering. But I found that the some of the cell identity IDs are skipped in the reclustered T-SNE plot (0,2,3,4,6... instead of 0,1,2,3,4,5,6):

enter image description here

Here is my code:

DefaultAssay(combined.all) <- "integrated"

# subset Seurat to only Epithelial cells
subset.Epithelial <- subset(x = combined.all, idents = "Epithelial cells")
# run tsNE
subset.Epithelial <- RunTSNE(object = subset.Epithelial)
# run umap
subset.Epithelial <- RunUMAP(subset.Epithelial, dims = 1:20) # reduction = "pca"

# T-sne plot
Epithelial.clusters.tsne <- DimPlot(subset.Epithelial, reduction = "tsne",group.by = "seurat_clusters",label = TRUE)

I cannot figure this out nor find help from other website resources. Could anybody provide some help on how to interpret this?

Thanks so much!

scrna

Now I know that I should have all the normalization steps before I RunTSNE() step. Now I have another question. So since I have already run the normalization step and the scale step on the entire Seurat object at the beginning, If I rerun them on this subset, will it run on raw counts or the normalized and scaled counts?

Thanks!

1 answer

The reason why you are missing some clusters is because you are not reclustering, and the column seurat_clusters in your final tsne still has the original clustering. What you need to do is to :

  1. Subset
  2. Run Umap/tsne
  3. Re-cluster ( FindNeighbors + FindClusters)

Thanks for your reply fracarb8! Just to make sure, so I don't need the normalization and scaling steps here? Cause there are other suggestions said I need to also run those, so I'm confused here.

Thanks! Leran

Log in to answer this question.