scRNA-seq: Consistent low number of cells and low fraction reads across the samples

Question

[newbee]

Hi,

cell ranger count returned the following representative result for a sample. But, this is consistent for all the samples in a project. The number of cells detected ranges from 19 to 35, and the fraction of reads in cells ranges from 13.1% to 58.1%. Is there any hope to move forward with downstream analyses with this data?

Any answer will be much appreciated. Thank you very much.

[rpolicastro]

Assuming Cell Ranger was ran correctly it's unlikely that the sample is useable. It looks like most droplets were probably empty.

[ATpoint]

Seconding that. Is there anything else that you can use for QC, like some Trypan Blue stain of the cells right before loading the chip? Is this an established protocol for that celltype or experimental? Are cells known to be fragile? As said above, the sample is likely unusable. You must now go through the sample preparation and try to find the reason why it failed.

[newbee]

Thanks for raising these valid points. I do not have the answer right now. Lately, I heard from the PI that they started with a small number of cells (less than 100 for some samples).

[GenoMax]

less than 100 for some samples

Hmm. That is surprising but then in light of that fact ~20% recovery of cells does not sound bad.

[rpolicastro]

They could have saved themselves a ton of money by sequencing this on a miseq.

[jv]

I wonder if using cell hashing, which would allow pooling of cells from multiple samples before droplet formation, could improve things here by increasing the total number of cells and thus reduce the empty droplet to cell droplet ratio.