Dot vs. Contour Plot

Data graphical display type is an important choice when performing a flow cytometry workflow. Effective data display enables the reader to accurately interpret displayed information.  The information collected by the flow cytometer is saved in a special type of list mode data (LMD) file, known as a flow cytometry standard (i.e. FCS) file. This information can be displayed within FlowJo™ by first binning the recorded data in a two dimensional histogram and graphing those bins on an x-y plane (cartesian coordinates). The axial parameters are usually some combination of light scatter and fluorescence (1). Two common graph types for flow cytometry data are dot and contour plots. Both graph types are bivariate (having two axes correspond to different variables). The optimal method of visualization can be selected based on the nature of the data set.

A dot plot is helpful when studying populations with moderate or low frequency populations. The data from each event (usually a cell) is binned into histograms, and the most highly populated of those bins are delineated on the plot as dots (1). Figure 1 is a dot plot illustrating scatter parameters, and 2k events displayed using up to 2k dots. 

Figure 1. Dot plot of side scatter vs. forward scatter (event count=2000)

In Figure 2, the same scatter parameters and population of 35k events is represented by up to 5k dots. 

Figure 2. Dot plot of side scatter vs. forward scatter (event count=35000).

When plotting data from populations with a large amount of variance, the dots may begin to bunch up. The bunching can obfuscate important details within the data (1). When the plot has concentrated high density regions, a contour plot may be more effectual for data presentation.

On a contour plot, the relative frequency of data are displayed. Event densities are depicted with contour lines (gradients). Each line encloses an equal percentage of events (1). Closely packed contour lines indicate a high concentration of events. The proportional nature of contour plots makes them more independent of sample size. Contour plot gradients can also be logarithmic, in which case each successive contour line contains twice the events as its precursor (1).  In Figure 3, the same 2k event population shown in Fig. 2 is graphed using a contour plot. 

Figure 3. Contour plot of side scatter vs. forward scatter (event count=2000). 

Events that fall beyond the outermost contour line are not represented by a standard contour plot (1). Analysis software permitting, a contour plot can include outlier events as dots. Figure 4 shows the 2k event population as a contour plot overlaid with outliers.

Figure 4. Contour plot with overlaid outliers (event count=2000).

Contour plots are also useful when there is a build-up of events along axes. Due to lack of discernible detail on dot plots, it may be difficult to approximate the number of events that are off-scale (1). Figure 5 is a side-by-side contrast of visualizations of the 2k population. The population has an accumulation of events along the positive end of both axes. 

  Figure 5. Contrasting plots of side scatter vs. front scatter with data accumulation (event count=2000).

Dot plots and contour plots are both helpful data visualization methods. Dot plots can be more advantageous when viewing rare populations. Contour plots are more effective with high event counts, because the relative frequency of events is displayed in more detail. Contour plots are also useful for viewing clustering around axes.

References 

1. Shapiro, Howard. Practical Flow Cytometry. New York, Alan R. Liss, 1985.