treekoR: An automated framework for elucidating hierarchical relationships in high dimensional cytometry data
Adam Chan,Jean Yang,Ellis Patrick The University of Sydney
High throughput single cell technologies which measure a high number of parameters for up to millions of cells holds the promise to discover novel biological relationships between different patient conditions with effective analytical workflows. Whether identifying cell clusters is done via unsupervised clustering, or through manually gating cell subsets, the proportions of these cell types relative to the whole sample are able to be analysed to understand characteristics driving different patient conditions. However, typical analyses employing unsupervised clustering techniques have overlooked incorporating the important hierarchical structure in the data through measuring proportions of cell subsets relative to a parent subset. We present treekoR, a novel framework that leverages hierarchies to recapitulate the significant proportions of cell subpopulations to parents. Our results over twelves case studies across three single cell technologies demonstrate the importance of measuring proportions to parents (along with absolute proportions) in the analyses of cytometry data that will lead to novel biological insight and improve patient outcomes.