Abstract
Quirks and All: unlocking comprehensive biological discovery out of your spectral flow cytometry datasets
David Rach1
Flow Cytometry Shared Resource, University of Maryland Greenebaum Comprehensive Cancer Center, Baltimore, USA
Spectral flow cytometry (SFC), with its capacity to resolve remarkably similar fluorophores while rapidly acquiring millions of events, has permitted the profiling of immune systems at breadth and depths unimaginable just a few years ago. For researchers working with limited biospecimen, this has been a game changer, allowing for wider understanding of human immune heterogeneity, and the response to infection and immunization.
While the analytical process remains essentially unchanged if only one cell subset is of interest, the increased dimensionality of the datasets compared to conventional flow cytometry necessitate the use of unsupervised or semi-supervised analytical approaches if researchers wish to characterize the majority of the cell populations within their acquired samples. However, many of these existing tools, originally developed for use with mass cytometry and single-cell RNA-seq, struggle due to both the increased number of events, as well as the inherent uncertainty in the unmixing process, with the resulting variation in MFI frequently introducing batch effects.
To address these unique quirks, over the last two years, we have been working on developing open-source tools in both R and Rust, facilitating troubleshooting of unmixing errors due to tandem degradation and additional autofluorescence; pre-emptive identification of issues arising from failures of instrument quality control; and enabling more thorough and reproducible analysis to profile all cell populations present within our datasets. We highlight these approaches in context of an on-going analysis of a rare clinical cohort of HIV-exposed uninfected (HEU) neonates, using 32-fluorophore SFC panel for innate-like and conventional T cells, acquired on a 5-laser Cytek Aurora. All the while, weighing the question: “What good is any tool if the community as a whole is unable to use it?”
License
In our commitment to open-science and open-source, all teaching materials are freely offered under a CC-BY-SA license, while all code examples are offered under the AGPL3-0 copyleft license.
