Cluster removal

Question

[fmazzio1]

Hello,

I am analyzing scRNAseq data and it seems to me that one of the clusters is sample-specific. I am comparing two conditions (responders and non-responders to therapy) and looking at my UMAPs by sample_ids, it seems that one of the clusters is present only in one patient. I have read papers where the authors removed clusters that were sample- instead of condition-specific. However, I was wondering if this should be done just looking at the clusters distribution onto the 2D UMAP or if there is a more unbiased approach that looks at the "unbalance" of clusters across the samples so that in case one cluster is clearly sample-specific can be justified the removal.

Thanks!

Best, Francesco