Enumeration of Low-Fluorescence-Positive Cells merged into a Shifting Background
Francis L. Battye
The Walter & Eliza Hall Institute
Fluorescent staining techniques aim to bind fluorochrome specifically to cells bearing the appropriate receptors while leaving all other cells untouched. The reality is that often all cells in a stained sample appear shifted in fluorescence intensity relative to the unstained "background distribution". If, after staining, there is no clearly separated group of stained cells but the fluorescence histogram becomes biphasic, with two apparent subpopulations merged together, then some computational assistance is required to extract the numbers of cells in each phase. The process is made more difficult if the background distribution cannot be overlaid against the stained cell distribution because of an apparent general shift.
A computational technique has been developed that is based on a reasonable model of the way the so-called background fluorescence might shift following staining. Assuming a distributed binding of fluorochrome to all cells regardless of specificity, the mean additional fluorescence is adjusted until the background histogram fits the lower component of the stained sample histogram. It is then subtracted, leaving the upper component for enumeration. Instrumental factors like electronic noise and imperfect analog logarithmic amplifiers have been taken into account. The fitting process has been found to be more robust when the lower component of the stained sample is in the majority. It is also improved for higher numbers of cells counted where the statistical accuracy is improved.
This technique is to be regarded as empirical and its success not reliant on the correctness of the model. Further, we refrain from categorically labeling the enumerated populations as "negative" and "positive", preferring the terms "low" and "high" fluorescence.
