Single cell analysis: Unable to subset cells in seurat object using desired nFeature_RNA cut off

Question

[sc_analysis]

#Violin plot to visualize QC matrices(before cell subsetting)
VlnPlot(object = merged_seurat_objects, 
        features = c("`nFeature_RNA`", 
                   "nCount_RNA", 
                   "percent_mt"), 
        ncol = 3)![enter image description here][1]


#Feature plots to visualize relationship between nCount_RNA & nFeature_RNA
 FeatureScatter(object = merged_seurat_objects,
                        feature1 = "nCount_RNA", 
                        feature2 = "nFeature_RNA")![enter image description here][2]


#Subset merged_seurat_objects_filtered(cell filtering)
merged_seurat_objects_filtered <- subset(x = merged_seurat_objects, 
                                         subset = nFeature_RNA > 200 & 
                                                  nFeature_RNA < 7500 &
                                                  nCount_RNA > 200 &
                                                  nCount_RNA < 5000 &
                                                  percent_mt < 10)
#Violin plot to visualize QC matrices(after cell subseting)
VlnPlot(object = merged_seurat_objects_filtered, 
        features = c("nFeature_RNA",
                   "nCount_RNA",
                   "percent_mt"),
        ncol = 3)

BUT...after subsetting, the violin plot is showing nFeature_RNA < 2000 as cut off. How to resolve this?

[fracarb8]

You are not subsetting based on nFeature. You are subsetting based on nFeature AND nCount AND perc mt.

[sc_analysis]

Yes. But after applying:

 merged_seurat_objects_filtered <- subset(x = merged_seurat_objects, subset = nFeature_RNA > 200 & nFeature_RNA < 7500 & nCount_RNA > 200 & nCount_RNA < 5000 & percent_mt < 10)

I am getting the plot where nfeature_RNA < 2000.

[fracarb8]

Looking at both nCount and MT the profile are cut at your thresholds, suggesting that you simply don't have cells with >2000(ish) genes. nFeature is inside the range you provided (>200 and <7500), so I don't see the issue. Are you expecting cells with 7000 genes?

[sc_analysis]

I am not sure what should be the cut off. Looking at the vlnplot before subsetting i thought most of the cells are falling under 7500 nfeature RNA . so i wanted to take it as cut off to avoid outlier cells that can be doublets. If i get take cells under 2000 nfeature_RNA then i may lose a lot off cells. AAs this is a general analysis i just wanted to set qc parameters. In other papers where they have used this dataset, they have used 5000-6000 for nfeature and n coubt RNA.