<- Back to Spectre home page

This page discusses various methods of axis or data transformation. For instructions to export these file types from FlowJo, please see this page.

Axis scaling with cytometry data

Excerpts on this page extracted from: Ashhurst, T. M.,Smith, A. L., &King, N. J. C.(2017).High‐dimensional fluorescence cytometry. Current Protocols in Immunology, 119, 5.8.1–5.8.38. doi: 10.1002/cpim.37

An important consideration for analysis is how data is viewed and interpreted. While fluorescence data was sometimes viewed on a linear scale, this was insufficient for simultaneously examining high and low fluorescence values. Because of this, the use of a logarithmic scale for visualizing fluorescence cytometry data became typical. However, viewing data on a typical logarithmic scale introduces misleading visual artifacts for signals at the low end of the scale. As such, alternative methods of plotting data have been developed, such as using a bi-exponential scale with Logicle constraints (Parks, Roederer, & Moore, 2006). In these Logicle transformations, the high end of the scale is logarithmic and the low end of the scale is converted into a linear scale. The scale then returns to logarithmic at values below the linear component.

Linear (max): The y- and x-axis have been plotted using a linear axis. In cytometry data, this is often not so useful, as a large number of the data points are crowded in the left hand side of the plot, with much of the space taken up by high-value data points and potential outliers.

Linear (modified max): If the linear plot is modified so that the maximum value is adjusted, the plot is slightly improved. However, much of the data is still crowded in the left hand side of the plot, and a number of events are off scale on the right hand side.

Logarithmic: A logarithmic plot improves the visualisation, by enabling a greater dynamic range – both low and high value data points can be effectively plotted on the one continuum. However, ‘0’ and negative numbers cannot be plotted on a logarithmic scale, these are stacked on the left hand side. Additionally, cells that are all ‘negative’ for the x-axis marker take up a large amount of the plot – this spread would be interpreted by a computational algorithm as relevant differences in expression.

Logicle/Bi-exponential: This plot is intended to be the best of both worlds. Most of the plot is logarithmic, allowing for an effective dynamic range to be plotted. The low end of the plot (left hand side) is compressed (reducing the spread of negative signal) and is switched briefly to linear around zero, allowing for the plotting of values of ‘0’, and plotting of negative values below 0.

The degree to which the low-end logarithmic scale is ‘compressed’ into the linear space is known as the ‘width basis.’ This method of scaling allows data with low values, as well as data with negative values, to be displayed more accurately. This approach is especially useful for visualizing data with high levels of SE, where the ‘negative’ or autofluorescent population spreads out substantially.