[Cytometry] Compensating when you don't have negatives

Mario Roederer roederer at drmr.com
Sat Feb 8 13:51:22 EST 2014

One more useful tidbit to add to all the good advice:

You can still compensate properly even if 100% of your cells are transfected (i.e., 100% GFP+).  Remember, for proper compensation, you don't actually need a "negative" control. You just need a positive that is at least as bright as any of your samples, and a "less bright" reference control. Even if the less bright cells are only tenfold dimmer than the positive (i.e. one decade left shifted) you should get accurate compensation estimation (note: the bigger this difference, the better, which is why we typically use unstained "negatives" as the reference).

What this means is that if you have 100% transfected cells but still have a good dynamic range of GFP expression (which is usually the case), one that ranges across at least one to two decades, then you could  set a gate on the brightest fraction of the cells in a separate gate on the dimmer fraction of cells; if these two case are separated by more than a decade you are in business ( I.e., you can use the dim date as your "negative" control).

One caveat: the rule that the autofluorescence distribution for the two controls should be the same if there were no GFP still holds. So: if the  brightest  GFP  transected cells were substantially different subset from the dimmer ones that had much different autofluorescence, you may  introduce some error.   on the other hand, if your "dimmer" population is much brighter than autofluorescence, then this error will be so small as to be meaningless and you will still get accurate compensation.

It's important to remember that the point of compensation is to estimate the slope between the two fluorescence channels --  and the slope can be calculated between any two populations that differ in fluorescence intensity --  the lower one does not have to be negative.

Note,  you may want to use this strategy even if you have a clear negative population: i.e. to date on the brightest fraction of GFP positive cells as your "positive" control.   since the estimation of compensation is  only accurate for intensities up to the intensity of the control population, you may get better accuracy by using the   brightest cells you have.  In fact, you may have noticed this: that while you used a GFP positive sample for your control,  the compensation is only accurate up to the median of that population, and any trailing cells that are 1 to 2 decades about this may appear under or overcompensated.

If you do use a sub-gate  of cells as your positive (or negative) control,  please remember David admonition below to use a gate that has at least 5 to 10,000 (and typically for cells I recommend 20 to 50,000) events.


On Feb 7, 2014, at 6:51 AM, David McDonald wrote:

> Just to add to Cris's message. If you have a very high transfection efficiency so that you don't have a negative histogram for your transfectants then either mix unstained cells into your transfectants or acquire and append  some unstained cells into the "GFP" compensation control tube (you need a minimum of 5000 events for the autocomp to work). 

More information about the Cytometry mailing list