High number of detected genes per cell after doublet removal - scRNAseq data

Question

[sarahmanderni]

Hi,

I have used doubletFinder to remove doublets from scRNAseq data but still a considerable proportion of cells have high nFeature_RNA (> 6000) while based on the quality control violin plots the median nFeature_RNA is between 2000 to 3000. Should I still remove the skewed cells in terms of number of detected genes after using doubletFinder?

[ATpoint]

Please show plots, or anything. Also, what is the experimental design? Are there many celltypes expected with very different numbers of expressed genes?

[sarahmanderni]

Here are the violin plots after removal of doublets. These are four samples from mouse lung tumor.

[sarahmanderni]

Also, could you please make an example about when there could be many cell types with very different number of expressed genes? do you mean differentiated and non-differentiated cell types?