S8 channel loss during unmixing

[DrCytometer]

In at least one data set, channels have disappeared when FACSDiscover S8 data are unmixed in version 1.5.1. The cause of this is unclear, but may be related to keywords.

[DrCytometer]

This is an FCS3.2 issue with certain BD-specific keywords. A fix is in progress.

[DrCytometer]

A patch for this has been posted in version 1.5.2. I'm leaving this issue open as there may be other keywords that could cause a similar problem.

[luis-crisostomo]

A patch for this has been posted in version 1.5.2. I'm leaving this issue open as there may be other keywords that could cause a similar problem.

The problem persists as of 1.5.3., observed with AutoSpectral and AutoSpectralRcpp using the "AutoSpectral" and the "FastPoisson" algorithms. Missing parameters:

• Sorted
• Eccentricity (all variants)
• Correlation (all variants)
• Saturated

AutoSpectral unmixed parameters are also missing from the unmixed file. Only 3 parameters were saved -> I added the tag "(AS-unmix)" after the marker in the csv control file to differentiate unmixed parameters from Chorus-unmixed parameters.
Unmixing command:
unmix.fcs(CTRL_male_pooled, output.dir = "./AutoSpectral_Unmixed_test", spectra = spectra_CTRL.males, asp = asp_s8, flow.control = flow.control_CTRL.males, method = "AutoSpectral", af.spectra = af.spectra_CTRL.males, spectra.variants = fluor.variants_12w_CTRL.males, include.imaging = TRUE, include.raw = FALSE, speed = "slow" )

[DrCytometer]

So, for the first part (Sorted, Eccentricity, etc.), these are intended to be excluded from the output files as they are imaging parameters. This is controlled by include.imaging in the arguments to unmix.fcs() and unmix.folder(). That argument is set to FALSE by default--shall I change it to TRUE? The intention here is make the files smaller and more manageable.

The second part may or may not be the same issue. AutoSpectral currently tries to locate existing parameters in the raw FCS file that match the new ones. This preserves FSC, SSC, Time, etc. In the case of the FACSDiscover systems, this may be matching to the default OLS unmixed channels already produced in Chorus. This could be tested by altering the fluorophore name in the CSV control file, rather than the marker, I think. For example: FITC2 instead of FITC. If that does not produce new unmixed channels, then it is the keywords.

[luis-crisostomo]

Thanks for the feedback. I was reporting the imaging parameters as missing even though I stated include.imaging = TRUE. But I must stand corrected - the parameters were saved, it seems that FlowJo loaded them in a completely wrong scale.

Regarding the missing unmixed parameters, I did 2 tests after the test with the "(AS-unmix)" tag. First, I run AutoSpectral() a S8 .fcs file exported in "Raw mode", i.e., without any unmixing matrix. The issue persisted, probably because the exported .fcs still includes marker parameters. Then, following your suggestion, I changed fluorochrome names in the control file such as "BV421 (2)". This solved the issue: all the unmixed parameters were added to the unmixed files, as intended.

[DrCytometer]

Okay, that sounds like something I should fix. Would you be able to send me an FCS file, the control file (less important) and the spectra CSV file (also less important)?