Seurat clustering Methods-resolution parameter explanation

Question

I am learning the Seurat algorithms to cluster the scRNA-seq datasets. I found this explanation, but am confused. Can someone explain it to me, "The FindClusters function implements the procedure, and contains a resolution parameter that sets the ‘granularity’ of the downstream clustering, with increased values leading to a greater number of clusters. We find that setting this parameter between 0.6-1.2 typically returns good results for single cell datasets of around 3K cells". My questions are:

1. What is "granularity" of the downstream analysis?
2. How does the parameter "resolution" work?
3. How to know the optimum value for parameter "resolution"?

I found this paper about the resolution that might help, but not clear to me yet. https://iopscience.iop.org/article/10.1088/1742-5468/2008/10/P10008/pdf.

I think the higher the resolution, the more clusters produced meaning the smaller clusters obtained? The smaller the clusters the more specific cell types we get, is that correct?

Answer 1

What is "granularity" of the downstream analysis?

Low granularity implies basic cell types (e.g. T cells, B cells), implies few larger clusters.

High granularity implies more refined cell types (e.g. CD4+ Central Memory T, Treg, Memory B cell etc.), implies more smaller clusters

How does the parameter "resolution" work?

A high resolution parameter value produces more smaller clusters, i.e. higher granularity. If you want to know how modularity clustering works, I recommend the Leiden algorithm paper https://www.nature.com/articles/s41598-019-41695-z

How to know the optimum value for parameter "resolution"?

Tricky, and it depends on what questions you want to answer. Normally it's a subjective iterative process of trying a clustering resolution, checking if the clusters make sense, re-trying etc. but I also quite like the clustree approach https://cran.r-project.org/web/packages/clustree/index.html

The smaller the clusters the more specific cell types we get, is that correct?

Yes correct