Tue Nov 11, 2014 11:49 pm by mleipold
HI Ofir,
I think there are a few topics contained in that question.
1. What cell populations are you looking at, and how rare are they?
-The Yao et al paper touches on this a bit: "Indeed, our study was not designed to identify the full spectrum of cells in each subset, but rather to determine the number of input cells necessary to identify cells labeling with a particular defined set of markers."
If you're just wanting to look at total CD4+, you'll need fewer cells.
2. What are you wanting to do with the cell data?
- Are you just counting them in your collection of however many cells? For instance, antigen-specific T cells are sometimes reported as count per 10 million cells.
- Are you wanting to actually calculate medians and such? If so, for statistical purposes, you'll probably need several tens to low hundreds of cells in order to have confidence in your medians, CVs, etc.
3. A couple years ago, we ran a similar experiment on our CyTOFv1. With even a bit more experience, my cell recovery would probably be a bit higher, but this was the trend that day.
"Cell #, plated" "Live intact Singlets"
200000 28315
400000 55035
500000 75524
750000 116182
1000000 174862
In summary, the 200K gave a recovery of 14%, 500K gave 15%, while the 1mil gave a recovery of 17%. Note: this includes the cell transmission efficiency loss of 70-80%, as well as all the losses with washing and other sample processing. If I'm reading Yao et al correctly, their numbers also reflect these sources of loss.
So, we at the HIMC generally ask our customers for at least 500K/sample, plated, in order to have a decent chance to go down to some smaller subsets (even just memory T cells), while ensuring half-decent statistics.
-Mike