a, Example heatmap of Spearman’s correlation statistics between 100 FlowSOM groups and Phenograph clusters based on median protein expression values, to illustrate the methodology for mapping FlowSOM groups to cell phenotypes assigned using Phenograph. Solid squares indicate largest positive correlations. Rows and columns are ordered by hierarchical clustering using Ward’s method (comparison based on n = 479, 844 cells). b, Bar chart on the left shows the number of cells assigned to the Phenograph cell phenotypes. Bar chart on the right depicts frequency of FlowSOM groups mapped to each Phenograph-clustered phenotypes, arising from 100 runs of FlowSOM each generating 100 groups (10,000 groups mapped in total). This illustrates that most phenotypes were assigned by both methods and that the frequency of mapped groups was related to the number of cells represented by each phenotype. c, Bar charts comparing patterns of association (tested using linear regression) with breast cancer molecular subtypes between mapped cell phenotypes (from 100 runs of FlowSOM each generating 100 groups as in panel b) and the Phenograph-assigned phenotypes. Green and red bars distinguish between enrichment and depletion of a given cell phenotype for each molecular subtype. ‘Original’ on the x-axis indicates associations based on the Phenograph clustering methodology.