Hi all,

I am using fastMMN for batch correction and subsequently buildSNNGraph and cluster_walktrap for clustering on single-cell rna seq data on multiple samples. However, I end up with way higher number of clusters compared to what I would expect. By setting the steps parameter in cluster_walktrap it is possible to change the number of clusters.

Are there any other parameters or ways to reduce the number of clusters identified, similar to i.e. setting the resolution in FindClusters in the seurat pipeline?

Best, Tobias

Increase the number of neighbours used (k) when running buildSNNGraph. This increases the connectivity of the SNN graph and should reduce the number of clusters (i.e., follow the opposite of the advice given in the comments section here). I am surprised that the default k=10 yields too many clusters, though; I can only assume you have loads of cells, such that 10 nearest neighbours becomes quite a small neighbourhood.