Question: I see a low value for the "Fraction Reads in Cells". How can I interpret this metric?
Answer: A low "Fraction Reads in Cells" value is typically explained by the following:
1) High ambient RNA (background) in your sample. This ambient RNA comes from lysed/dead cells in your sample. Cell Ranger is able to confidently align the reads from ambient RNA to the transcriptome but the reads are not associated with a valid cell-containing GEM.
2) The cell-calling heuristic did not apply. For example, there may be higher variation in RNA content than expected (more cells with lower RNA content). The current cell-calling heuristic assumes a ten-fold variation in RNA content.
Cell Ranger's algorithm for partitioning barcodes as cells versus background is based on the idea that barcodes for cells should have distinctly more transcript counts associated with them than the background barcodes. This can be visualized by the ranked barcode plot in the web_summary.html file. More details on the cell filtering algorithm can be found here and here.
If you suspect that Cell Ranger's cell calling algorithm did not work well for your sample, please re-run
cellranger count again or
cellranger reanalyze with
--force-cells option to call the expected number of cells.