Cellranger counts: how to set the value for "--expect-cells" if you can't guess

Question

[Aaron]

I see "cellranger counts" recommends people estimate the number of cells in the samples they want to analyze (with the "--expect-cells" argument, and it has a default value of 3000 cells if you don't input a value) - if I don't know how many cells are in my sample and I couldn't make a good guess, is there any good guidance for what to do then? I haven't seen any questions on Biostars or elsewhere about this in particular.

Cellranger counts: https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/using/count#cr-count

[GenoMax]

You can start with the default and see what you get. If you had counted the number of cells that went into the experiment then you could try that number.

[Aaron]

Thanks GenoMax this is very helpful. Unfortunately, I'm using data collected by someone else and I can't seem to find any record of cells counted - is there a way to estimate the number of cells if they weren't initially counted?

Answer 1

[shiyang_bio]

Generally speaking, using 3000 or 5000 won't generate too much difference. But 300 and 5000 may lead to significantly different result. Cellranger just use this value as a rough range for its algorithm. The final cell number is determined by the "knee plot".

[Aaron]

This is very helpful @shiyang_bio, thank you!