A Practical Strategy for Analysis of Large Cytometry Data through Supercells

Author(s): Givanna H. Putri, George Howitt, Felix Marsh-Wakefield, Thomas Ashhurst, Belinda Phipson

Affiliation(s): Walter and Eliza Hall Institute of Medical Research



The rapid advancements in cytometry technologies have enabled the quantification of up to 50 proteins across millions of cells at a single-cell resolution. The analysis of cytometry data necessitates the use of computational tools for tasks such as data integration, clustering, and dimensionality reduction. While numerous computational methods exist in the cytometry and single-cell RNA sequencing (scRNAseq) fields, many are hindered by extensive run times when processing large cytometry data containing millions of cells. Existing solutions, such as random subsampling, often prove inadequate as they risk excluding small, rare cell subsets. To address this, we propose a practical strategy that builds on the SuperCell framework from the scRNAseq field. The supercell concept involves grouping single cells with a very similar transcriptomic profile, and has been shown to be an effective unit of analysis for scRNAseq data. We demonstrate the effectiveness of our approach by conducting a series of downstream analyses on six publicly available cytometry datasets at the supercell level, and successfully replicating previous findings performed at the single cell level. We present a computationally efficient solution for transferring cell type labels from single-cell multiomics data which combines RNA sequencing with protein measurements, to a cytometry dataset, allowing for more precise cell type annotations. Our enhanced SuperCell framework are available on our GitHub repositories (https://github.com/phipsonlab/SuperCellCyto).