Seurat cluster is split and appears in 2 different areas on the tsne plot

Question

[asmariyaz23]

Hello,

I am analyzing a single cell dataset using Seurat. The tsne plot obtained shows the same cluster in 2 different locations. I am wondering if I should be merging them or play with the parameters for tsne to separate them out further. To study these 2 clusters I want to look into the genes in these 2 locations. Any idea how I could do this?

Thank you, Asma

[i.sudbery]

Its always sensible to treat tSNE as a visualisation only tool, and not to use tSNE results for inference. My guess would be that if you are clustering and finding that your clustering algo treats something as one cluster, but it looks like two on your tSNE, that your clustering is right and the tSNE wrong. Thus I would play with the tSNE parameters until it looks like one cluster.

Answer 1

[asmariyaz23]

I have been trying to play with tsne parameters as you suggested, mainly max.iter, min_dist and n_neighbors. For some reason it is the 13th cluster which splits up and occurs in 3 different locations. Image in link below. Is there some other parameter that I should explore?

https://ibb.co/YPH1pnV