Benchmark for resolution parameter in Seurat::FindClusters()

[AleksandarMih]

Hi,
I'm getting started with Seurat, and I'm currently attempting to cluster the cells of a dataset with 33,000 cells distributed across 18 patients. I downloaded the dataset from an existing paper where the author ends up with 91 different cell types. How should I choose the resolution in this case? Are there any general benchmarks regarding the number of cell types and the total number of cells that can help narrow down the search for the optimal resolution parameter to a given interval? Thanks!

[EZUY]

Resolution controls how fine the clustering is — higher means more clusters. For a dataset this size (33k cells, 91 reported types), I’d start around 2.0 and adjust up(likely up; too many clusters)/down while checking marker genes and biological interpretability. Often you do broad clustering first, then subcluster lineages for finer resolution.

[AleksandarMih]

Thank you very much!