[Help] Recreating plots from FCM using FlowKit

[TobiasBR]

Hi! I'm working with the flow cytometer (FCM) called "BactoSense CORE" by bNovate in my master thesis. I'm trying to recreate the plots produced by machine using raw data and FlowKit. When applying the Logicle transform, it maps scale values to [0,1], which does not match the plots from the FCM. Is it possible to avoid this scaling or inverse it so the FlowKit plots matches with the BactoSense one? Any help would be appreciated!

Attached are screenshots of each plot.

[whitews]

Hi Tobias,

Thanks for using FlowKit. The short answer is that FlowKit displays the axis values for the transformed space because you need to reference those values for gate boundaries when using transformed data. There's a more detailed explanation here:

#195 (comment)

Hopes this helps to at least explain the difference you are seeing. At some point in the future I may add a plot option to display the ranges in the non-transformed space, but the current version doesn't have this option.

Kind regards,
Scott

[TobiasBR]

Hi Scott,

Thanks for your explanation! That makes sense. I managed to recreate the gating from the BactoSense by creating a data frame with the gate vertices as shown in the image, converting it to a sample, applying the Logicle transform, and then convert it back to a data frame to get the right values for the transformed space. So the gating should be correct now.

Even though it's not a feature in FlowKit to display the ranges in the non-transformed space, do you know if it would be possible to "fake" this when plotting in Bokeh?

Cheers,
Tobias

[whitews]

I haven't researched it fully, but my guess is that Bokeh is sufficiently flexible to allow remapping the axis labels. I did make the design decision to return Bokeh plot objects in plot methods and not just show the plot, to allow users to modify the Bokeh plot.

If you're willing to give it a try, I'd be interested in seeing a prototype example demonstrating how to do this. After some googling, I made a very simple example:

from bokeh.plotting import figure, show
from bokeh.models.formatters import CustomJSTickFormatter

# Sample data
x = np.arange(10, 1001, 10)  # original linear values
x_log = np.log10(x)
y = np.arange(1, len(x) + 1)

# Create the plot with a log scale on the x-axis
plot = figure(x_axis_type="log", title="Log Scale with Linear Labels")

# Add a line glyph
plot.line(x_log, y)

# Define the JavaScript formatting function
formatter_code = """
    return Math.pow(10, tick).toFixed(2);
"""

# Create a formatter
formatter = CustomJSTickFormatter(code=formatter_code)

# Apply the formatter to the x-axis
plot.xaxis.formatter = formatter

# Show the plot
show(plot)

This uses the CustomJSTickFormatter which won't work so easily for the actual flow transformations, but does show the concept. Ideally we would be able to use Python functions to use the inverse transformations already in FlowKit. Again, any help in fleshing this out further would be much appreciated.

-Scott