I have 10x Genomics single-cell RNA-seq data from collaborators that I processed with CellRanger 5.0.1. All the data metrics look ok (98% valid barcodes, 77% reads confidently mapped to the genome, total genes detected as expected, knee plot ok, etc.), except for the number of cells detected. We expected 2000 cells, and got less than 200. However, the total number of reads is fine, so we end up with a mean of more than 1 million reads per cell.
I wondered if some of you would have experienced the same kind of issue or/and found what could have caused this cell loss. Could it be a bioinformatics problem (ie cellbarcode identification) ? An experimental one ? My guess is that cells did not separate well, so in each droplet we end up with 2 or more cells. In the end we count the number of droplets (so much less than the "real" number of cells) but we recover the right number of total reads.
Any advice is welcome !
Assuming Cell Ranger worked correctly, it likely means few cells were actually captured by the Chromium, and those few cells were over-sequenced by a factor of 100. Since it's likely an experimental error, you would need to consult with the people that loaded the cells.