AutoSpectral: OMIP-102 examples

Most of the examples in the AutoSpectral pre-print come from a pretty lousy dataset that was designed for the Babraham Spectral Symposium. This was for a workshop on finding and fixing unmixing errors, so it contains a bunch of marker:fluorophore combinations that are tricky and all of the fluorophores that emit where autofluorescence tends to be are included.

What happens if we run AutoSpectral on a good dataset, with cell-based controls? Here are bunch of examples from OMIP-102. You can download the FCS files and have a go yourself here.

All the plots here are gated just on FSC/SSC-Violet, then single cells by FSC-A/FSC-H and SSC-A/SSC-H. I've tried to ensure that the biexponential scaling is identical--let me know if I've missed a spot (in some cases, FlowJo is displaying the final 10^7 or 10^6 label and in others it isn't, and I don't know why). I'm also displaying a bunch of combinations that I expect may be problematic, either because of unmixing spread, spillover spread or tandem breakdown.

Correction: The original version of this post listed the original unmixing for OMIP-102 as being OLS. This was not correct. Thanks to the folks at BD, in particular Peter Mage, for catching this. The unmixed channels shown here as the originals are the SpectralFX channels, which is to say, the hard-coded channels in the FCS files. There are two sets of unmixed channels in data from the FACSDiscoverS8 and A8: one with plain labels (the SpectralFX channels) and one with with "SW-unmix", which is a version created in FlowJo using the spillover matrix from Chorus. SpectralFX is a proprietary algorithm from BD meant to reduce spillover spread in the unmixed data by taking into account Poisson noise from the detectors. The upshot is the SpectralFX should provide better unmixing than OLS. Check out the files on Mendeley or on the original OMIP-102 FlowRepository if you'd like to see the difference between the unmixing options offered by Chorus and FlowJo. You'll probably need FlowJo v11 to use the SW-unmix channels properly, and in FlowJo v10, these will just be OLS unmixing.

Please correct me further if any of this is wrong.

Red-headed Barbet, Costa Rica