Cell ranger reads do not match Seurat reads

Question

[firestar]

This is the summary from the CellRanger run showing ~150,000 reads per cell.

But, when I read the results into R, I get only ~5000 reads per cell. This is before any pre-filtering in Seurat.

x <- Seurat::Read10X_h5("data/PBMC_1_3_1/count/sample_feature_bc_matrix.h5")$`Gene Expression`
hist(colSums(x))

median(colSums(x))
[1] 5050.5

This is a huge difference and I wonder if I am making a mistake somewhere.

Answer 1

[rpolicastro]

Your Seurat histogram is actually of unique UMIs per cell, not raw read counts per cell. If you look at the cell ranger results under "Median UMI counts per cell" it matches (almost exactly) to your results from Seurat.

[firestar]

Ah right! I knew I was overlooking something obvious. Thanks!

[swbarnes2]

Yeah, 10xGenomics cheats a little by displaying the total reads far more prominently than the value you really want, the # of UMIs.