Statistics in Flow Cytometry Data and "MFI" values

                                The speed of the flow cytometry offers wide range of data points and data plots. Due to its sensitivity and versatility, it has been used extensively. Flow cytometer can detect up to 1000s of cells per second.. But, it should be noted that the flow cytometry also involves the statistics, its significance, calculating the fluorescence intensity.. 

              Statistics in flow cytometry involves, total number of data points acquired, percentage population, most importantly mean and median fluorescence intensity, and others. Most of us get confused about the term "MFI", whether it means mean fluorescence intensity or median fluorescence intensity.. Truth is, both.. It means it can be described as either mean or median fluorescence intensity. Some researchers use Mean for MFI and some use Median for MFI values. 

(Keeping in mind that, consideration of statistics in flow cyomtery also depends on the type of Application you are interested in. At the end of this article, you can read some of the examples)

How does Statistics Table Look like in Flow Cytometry Data?
                            Here is an sample image..                     

In the above statistics table, you can see that there are number of events, percentage total population, percentage parent population, mean, median and mode values for the specific detector. 

Observe difference in mean, median and mode values for the gate P2. 

What is MFI and When to use it?                     

                             MFI is usually used as mean, median or mode fluorescence intensity.. most of the times, its mean or median fluorescence intensity.. For most of the researches, its always confusing whether MFI should be described as a mean or median fluorescence intensity.. Well, it depends, you can consider either mean or median until and unless it is asked by the reviewer which one to use. 

Consider you choose, mean for the MFI, but, fluorescent intensity increases logarithmically, when it is logarithmic, then mean is useless to understand the population of events as the right-hand skew causes more exaggeration of the mean.
To overcome the problem in considering the mean, geometric mean is often used but, even geometric mean fluorescence intensity is susceptible to significant shifts. 

So finally the "median", median is less influenced by the skews and outliers in the data, and it is considered as more robust statistics.

To have more significant and robust MFI values, it is safe to use the median values in the statistical analysis. Because median is the mid point of the population (middle channel) and less influenced by skews and outliers.
MFI values should be used only if you think that it is best way for the comparison. Otherwise, you can just use the clean histogram plots and overlay histograms. and clusters in dual parameter plots. 



Let me give you three examples of how, when and which statistical parameters to be considered for different applications in flow cytometry. 
Examples:

1) Suppose, you are interested in "Fluorescence Shift" of the fluorescence of your interest and you would like to calculate the intensity change or fold difference. 
So, in this case, just make sure that which one is the best to choose to calculate fluorescence intensity. I would suggest you to use, median values as it is less influenced by skews and outliers in the histogram plot. 

2) Suppose, you are performing an experiment which involves the studying the apoptosis of cells.. That means "Apoptosis Analysis".. Now, you are only interested in percentage populations. (not in mean or median, they are not required here).. Here, only thing you have to worry about is percentage population of live cells, apoptotic cells or dead cells. 

3) Suppose, you are preparing the samples for "Cell Cylce analysis". So, in this case, you are interested in percentage populations of G0/G1 phase, S-phase and G2+M phase, some researchers consider the percentage population of Sub-G1 phase also..  and some researchers are also interested in the mean or median values of those peaks.. But, mean and median values only for some specific applications. But, generally you can just consider the percentage populations in each those gated histogram peaks.