Picking optimal resolution for single cell in seurat pipeline

Question

[synat.keam]

Dear all fellow members,

As I progressed into single cell analysis, one question that I would like to ask is how do we know the optimal resolution we should pick for our data as cluster will change once the resolution change. Not sure how to know?

Thanks

Tcell.cluster_UMAP<-  RunHarmony(Tcell.cluster, group.by.vars = "orig.ident")
Tcell.cluster_UMAP <- FindNeighbors(Tcell.cluster_UMAP, reduction= "harmony", dims = 1:40)
Tcell.cluster_UMAP <- FindClusters(Tcell.cluster_UMAP, resolution = 0.8)
Tcell.cluster_UMAP <- RunUMAP(Tcell.cluster_UMAP, reduction = "harmony", dims = 1:40)

Answer 1

[Ming Tommy Tang]

Take a look at this tool https://github.com/BradBalderson/Cytocipher I also have a blog post https://divingintogeneticsandgenomics.com/post/scrnaseq-clustering-significant-test-an-unsolvable-problem/

Answer 2

[ATpoint]

There is no answer to this, as it depends on your analysis. Choose the default one and see whether the clustering is fine to answer your biological question. If it turns out that a cluster might harbor two celltypes that you want separated then either change resolution and repeat or subcluster this larger cluster. I do not believe that there is one single, reproducible and universal way of predicting the "optimal" parameters for anything. If your goal is novel celltypes or states you naturally increase resolution. If you're more after general pickup of very different types of cells you leave resolution a bit lower to avoid overly many clusters. It all really depends, in scRNA-seq even more than in other assays.