How many cells do you need to reliably detect a population of interest?

The answer depends on the confidence you need, and the variability you can accept. Variability is expressed by the coefficient of variation (CV) is simply the standard deviation divided by the mean. The higher this number, the more “variable” the measurement. The lower the number, the less “variable” the measurement. Intuitively, a population that appears a lot, say 10% of the time, needs fewer cells than a population that occurs 0.001% of the time.

For purposes of this answer, let’s consider two populations, intermediate Monocytes (inMono) and T cells. These figures are derived from our mass cytometry validation report: https://teiko.bio/technology/, however we believe these reference ranges are applicable regardless of the instrument type (spectral cytometry or mass cytometry).

Cell Population	Median	Inter-run CV%	Intra-run CV%
inMono	0.47%	16.09%	4.63%
T cells	37.28%	0.39%	1.27%

Luckily, this question has been addressed in the work of Keeney et al. To wit, “for cell-based assays such as flow cytometry, a simple calculation can be used to determine the size of the database/sample that will provide a given precision: r = (100/CV)²; where r is the number of events meeting the required criterion, and CV is the coefficient of variation of a known positive control.”

We’ve adapted Keeney’s table below:

Desired Coefficient of Variation (%)		1	5	10	20
r = number of events of interest		10,000	400	100	25
When occurring at a frequency of:
Fraction	1:n	Total number of events which must be collected
0.1	10	100,000	4,000	1,000	250
0.01	100	1,000,000	40,000	10,000	2,500
0.001	1,000	10,000,000	400,000	100,000	25,000
0.0001	10,000	100,000,000	4,000,000	1,000,000	250,000
0.00001	100,000	1,000,000,000	40,000,000	10,000,000	2,500,000
0.000001	1,000,000	10,000,000,000	400,000,000	100,000,000	25,000,000

Now, let’s couple that with the instruments we have at our disposal: mass cytometry and spectral flow cytometry. Based on our experience, mass cytometry and spectral flow have recovery rates of 50% and 90%, respectively.

Drawn from Patient (mL)	Cells per mL (M)	Number of Cells in a Vial (M)	Instrument	Recovery Rate	Resulting Events (M)
3	1.8	5	Mass Cytometry	50%	2.63
			Spectral Flow	90%	4.73

Drawn from Patient (mL)	Cells per mL (M)	Number of Cells in Whole Blood (M)	% Granulocytes	Approximate number of Granulocytes (M)	Non-Granulocytes (M)	Instrument	Recovery Rate	Resulting Events (M)	Resulting non-granulocyte events (M)
2	5	10	50%	5	5	Mass Cytometry	50%	5.00	2.50
				5	5	Spectral Flow	90%	9.00	4.50

For PBMCs, you can get 0.5M – 10.8M events, depending on the volume of blood collected. In the table, we just show the average ranges, i.e. 2.63M – 4.73M. And for whole blood, you can get 4M – 10.8M total events. Since at least 50% of the cells will be granulocytes, you’ll get to ~2M – ~5.4M non-granulocytes. For readability, we’re showing only a smaller range, from 2.5M – 4.5M.

Now, let’s come back to the two populations from Table 1, inMono and T-cells.

Desired CV	1%	5%
Teiko observed inMono Median % of non-Granulocytes	0.47%
Total Number of events that must be collected, based on Keeney Table	1,000,000	40,000
Estimated inMono Population needed to achieve CV	4,700	188
Actual Teiko collected inMono events	302
Above Threshold?	No	Yes
Actual Teiko Intra-Run CV	4.63%
Actual Teiko Inter-Run CV	16.09%
Industry Standard CV Acceptance Criteria	25-30%

For a 5% CV for a 0.47% population, you would need about 188 events to achieve this CV. And We collected 302, so we were above the threshold. Turns out, we were right in line with inter-run CVs, at 4.63%, and Inter-run CVs at 16.09%.

Desired CV	1%	5%
Teiko observed T-cell Median % of non-Granulocytes	37.28%
Total Number of events that must be collected, based on Keeney Table	100,000	4,000
Estimated T-cell Population needed to achieve CV	37,280	1,491
Actual Teiko collected T-cell events	39,210
Above Threshold?	Yes	Yes
Actual Teiko Intra-Run CV	1.27%
Actual Teiko Inter-Run CV	0.39%
Industry Standard CV Acceptance Criteria	25-30%

Now, T-cells are much more populous. In our precision study, we collected 39,210 events, beyond the 37,280 and 1,491 cells we need to achieve a good coefficient of variation. And, as it turns out, we clocked 1.27% CVs for intra-run and 0.39% CVs for inter-run.

So, hopefully this gives you an intuition for “How many cells do you need to reliably detect a population of interest?” Interested in capturing a specific population?

Contact us to discuss further!