I did scRNAseq on the Chromium Flex Assay platform and I'm having issues with one of the samples that was run. For more context we are sequencing FFPE samples from splenic tumors that are in the transition state from a polyclonal into a more aggressive monoclonal tumor. For one of these samples, I'm getting one cluster that doesn't seem to be very informative and I'm starting to think it's representing some form of doublet population. This is because the cells of this cluster are split across many different other clusters on the UMAP, and when I look at the markers a high percentage of cells express erythrocyte markers at a low level followed by a mix of other cell markers. I've tried running DoubletFinder but it's failing to identify all of the cells in this cluster as doublets, so when I subset and recluster the population is still there. The cluster I'm referring to on the UMAP below is cluster 1 (sorry the label is overlapped with cluster 13).
For reference: 0 - B Cells, 2 - T Cells, 3 and 9 - Erythrocytes
I guess my issue is I'm nervous to fully subset an entire (fairly large) cluster outright in case I'm removing true biological cells. How should I go about handling this situation?
Here are the cell markers from running FindAllMarkers in Seurat:
I also see a couple of mitochondrial genes in the cluster. Did you perform all the initial QC steps to remove low-quality data?
Hi, yeah standard QC steps, I filtered out all cells with a percent.mt greater than 10%.
I would recommend performing an additional filtering step for cells with a high quantity of haemoglobin genes.