[Cytometry] Compensation basics

Dmitri Lubenets lubenets at ut.ee
Wed Jan 4 10:48:31 EST 2012

Dear all, 

in the light of on-going conversation about compensation I would also like
to ask a theoretical, but very confusing  question. Please accept my
apologies for a long text, I really don't know how to explain my question
shortly. I wonder how computer calculates proper compensation in a situation
when particle is stained with two fluorochromes "A" and "B" and both
fluorochromes send their signal into two channels (like FITC & PE
combination)? On my opinion to calculate proper compensation the one must
know "true" (it means already compensated) number in a first place. I'll try
to explain what I mean. Assume that you have a particle stained with
fluorochromes "A" and "B". After doing compensation controls with only one
dye at a time you find out that "A" spills over into "B" for 20% and "B"
spills over into "A" for 10%. After running double positive (AB positive)
particle through flow cytometer you'll get a non-compensated data where "A"
signal equals to 3000 and "B" signal equals to 10000. Now, when you start to
compensate calculations should be as follows:

20% of 3000 is 600, so "true" (I mean compensated) signal of "B" is

10% of 10000 is 1000, so "true" signal of "A" is 3000-1000=2000. 

However, the number of 3000 contains signal spilled over from "B", so it is
incorrect to take it for calculation. It would be correct to take 2000 and
calculate 20% from 2000, as only this number represents the "true" signal
part of which is spilled over. In this case 20% of 2000 would be 400 and
"true" signal of "B" would be 10000-400=9600. I agree that difference
between 9600 and 9400 is little, but if you imagine compensation matrix with
"A" spilling over into "B" for 200% or even 400%! Taking 400% from 3000 or
2000 makes a big difference. 

Please point out if you find any fundamental mistake in my calculations. 

Thank you in advance, Dmitri Lubenets


More information about the Cytometry mailing list