Fig. 2

Measuring %parent can provide additional insight over %total. a Scatterplot of test statistics with the cell clusters in differentiating between latent CMV infection patients. Highlighted clusters are significant using %parent, while not significant using %total. b Comparative boxplot of the proportions of highlighted cell clusters, between patients with CMV and without CMV, with the %total (upper panel) and %parent (lower panel). c A heatmap generated using treekoR on a CD8+ T cell compartment to predict healthy vs COVID-19, containing a hierarchical tree of cell clusters colored by the test statistic using the corresponding %total (nodes) and %parent (branches). The heatmap is colored by the scaled cluster median expression values characterize leaf nodes in the tree. d Scatterplot of test statistics of each cell cluster with test statistic from using %total (x-axis) vs. test statistic from %parent (y-axis). The HLA-DR+ CD38+ cluster highlighted has a larger test statistic when differentiating between COVID-19 patients and healthy control using %parent than %total. e Comparison of − log10 of p values of a HLA-DR+ CD38+ subset for %total, %parent, and manually gated proportions from De Biasi et al. from a t-test (pink) and Wilcoxon test (green). f Comparative boxplot of a HLA-DR+ CD38+ subset, with the %total (upper left panel), %parent (lower panel), and manually gated proportions (upper right panel) between COVID-19 and healthy patients