What is MFI and how is it calculated?

MFI is typically understood as mean fluorescence intensity. However, it is important to know which kind of mean we are talking about.

1. Median: midpoint of population (middle channel). Preferred method to measure MFI of a logarithmic histogram.

2. Arithmetic mean: number of events in each fluorescent channel divided by the number of channels. Because fluorescent intensity increases logarithmically (and most flow data are logarithmic), arithmetic mean quickly becomes useless to generalize a population of events, as a right-hand skew causes even more exaggeration of the mean and accurate MFI measurements cannot be made of events off scale either at the negative/dim or bright ends of the histogram.

3. Geometric mean: Without going into too much mathematical detail, the geometric mean compensates for that and is considered the second best choice of describing the MFI of a logarithmic histogram.

4. Mode: refers to the channels which are most frequented.

In an even distribution (example A) the median, arithmetic mean, mode and the geometric mean is almost identical. However, if we look at example B, a skewed antigen expression causes the mean to drift in the direction of the skewed area (in this case to the left). In order to compensate for this, the geometric mean (gMFI) is often used to account for the log-normal behavior of flow data. The MFI should NOT be used in a bimodal distribution (example C) as any average only holds true for normal distributions, and a bi-modal population is by definition not normal. Gating each population and presenting percentages will provide much more useful information.



Although the MFI is often used to define and describe the mean intensity and level of antibody expression, it should be noted that this assumes that the instrument is optimized including the voltage and compensation settings. Fluorescent intensity is sensitive to experimental condition (e.g. voltage, compensation, antibody dilution, tandem dye degradation, laser fluctuations, etc.), it can be misleading when comparing intensity of any kind across multiple experiments. A more meaningful way may be to measure the signal to noise ratio (mean of an antigen-positive population / mean of antigen negative population) while no compensation setting are applied. If compensation settings are applied, the MFI's of the antigen-negative as well as the antigen-positive populations are often affected. Looking at the example B, this would result in a S/N ratio of 77:1 for the CD7. This is typically done for antibody titrations and a titration curve is used to determine which signal/noise ratio is optimal for the antigen.



Conclusion: The experts recommend using the median (preferred) or the geometric mean (second best choice) for the evaluation of MFI on a logarithmic scale.