Fluorescent cell barcoding (FCB) enables efficient collection of tens to hundreds of flow cytometry samples by covalently marking cells with varying concentration of spectrally distinct dyes. A key consideration in FCB is to balance the density of dye barcodes, the complexity of cells in the sample, and the desired accuracy of the debarcoding. Unfortunately, barcoding bench and computational methods have not benefited from the high dimensional revolution in cytometry due to a lack of automated computational tools that effectively balance these common cytometry needs. DebarcodeR addresses these unmet needs by providing a framework for computational debarcoding augmented by improvements to experimental methods. Adaptive regression modeling accounted for differential dye uptake between different cell types and Gaussian mixture modeling provided a robust method to probabilistically assign cells to samples. Assignment tolerance parameters are available to allow users to balance high cell recovery with accurate assignments. Improvements to experimental methods include: (1) inclusion of an "external standard" control where a pool of all cells was stained a single level of each barcoding dyes and (2) an "internal standard" where each cell is stained with a single level of a separate dye. DebarcodeR significantly improved speed, accuracy, and reproducibility of FCB while avoiding selective loss of unusual cell subsets when debarcoding microtiter plates of cell lines and heterogenous mixtures of primary cells. DebarcodeR is available on Github as an R package that works with flowCore and Cytoverse packages at github.com/cytolab/DebarcodeR.
Keywords: algorithm; automated analysis; fluorescent cell barcoding; multiplexing; software.
© 2021 International Society for Advancement of Cytometry.