Objective:
To familiarise the user with how the various plot options can change the emphasis on the data displayed, and how this can be utilised when preparing graphics for publication.

Different software packages display the saved data at different resolutions. Some packages have the ability for the user to select the display resolution, while in others the resolution is fixed. Consider the plots below, which show the same data file displayed with 64,128 & 256 channel resolutions and various plot options. (This is a large active graphic.. Please wait while it loads)

(plots created in WinMDI version 2.8).
You can view the same options but plotted with Flowjo here, or similar display with FSC Express,
So what do the plot options mean and why is it important that you have an understanding of the affect a particular option will have?

Firstly, let's consider the data resolution issue as the "problem" we see here is intimately related to the display method choosen for a given data resolution.
Click here to see a WinMDI plot showing the use of 256 channel data resolution to "hide" a 71 event population, which is evident in a 64 channel resolution plots with all other plot formatting options set identically. The explanation for this is that each event in a 1024 channel data file gets its raw value divided by 16 and is subsequently packed into a 64 channel array (improving the chances of "seeing" rare events with a constant threshold of 1), whereas at 256 channel resolution one fourth as many channels get packed into a single channel. Therefore, the default display threshold of 1 is four times as likely to show a rare population at 64 x 64 than at 256 x256. If any smoothing is done, some array elements may contain values less than 1, so the threshold should be lowered to 0.5 or 0.1 as dictated by the data set and the number of smoothing iterations. If your interest is in elucidating rare populations of cells, then a thorough understanding of the interaction of display resolution and threshold is essential. That said, how many of you really understand the interplay of options that you are currently using?

Precis of the formatting options for contour plots:

Probability Contour Plots:
One of the most commonly used formats for contour plots, this format draws contours as a percentage of the total number of events. Therefore the area between each pair of contour lines contains the same percentage of the total events.
Linear Density Contour Plots:
This format draws contour lines as a percentage of the maximum event number with equal spacing between the lines. Therefore the "peaks" which represent the largest number of events tend to be emphasised, and lower populations of cells not shown.
Log Density Contour Plots:
This format draws contour lines as a percentage of the maximum event number with logarithmic spacing between the contour lines. This method can provide a good combination of detail at lower event numbers, while still showing the high numbers of events in peaks.

In light of the above discussion, for publication purposes, the choice of resolution for data display is critical. Many publications do not state what method of data analysis/display has been used yet small sub populations of unwanted events can be completely hidden by the appropriate, or inappropriate choice of contour plot options. This is one reason why dotplots are popular, however be aware that what size the final reproduction in the paper or journal is before deciding to forsake the contour plot. A contour plot with low channel resolution, or higher channel resolution and a smoothing algorithm applied often display data very clearly when reproduced for publication.
One display option that deserves special mention is the combination of contour or density plot, with dots displayed below the lowest contour line. This formatting option removes the above caveat , and increases the usefulness of displaying data using contour plots.

Finally, for publication purposes, plot the graphics which will appear in your paper at the publication size. Convince yourself and others, that these graphics are clear, representative of the data and serve to highlight the points which are discussed. If not, consider whether the publication would be as convincing if it lacked the visual impact and the flow cytometry data was presented in a tabulated form, say as relative median fluorescence intensity values.

Click here to go back to Histogram discussion:

or be a "glutton for punishment" and

Click here to go back to Regions and Gates.

FACS HOME