Fig. 1

A and B give two views of the topic model fit to the PBMC data [29] (n = 94,655 cells, K = 6 topics) using Structure plots [70, 71]. Cells are arranged horizontally; bar heights correspond to cell membership proportions. In A, the cells are arranged using the estimated membership proportions only. In B, the cells are grouped by the FACS labels (the “T cells” label combines all sorted T cell populations other than CD8+ cytotoxic T cells). In C, the topics are annotated by distinctive genes from the GoM DE analysis (Fig. 3) and by enriched gene sets. Numbers in parentheses next to genes give posterior mean l.e. LFCs, and for gene sets, they are enrichment coefficients. An enrichment coefficient is an estimate of the expected increase in the LFC for genes that belong to the gene set relative to genes that do not belong to the gene set. Note the groupings a–g in A are intended only to aid visualization. See also Additional file 1: Fig. S1 for an alternative visualization