Fig. 5: Distinct cellular states of NK cells identified in pan-cancer atlas.

a, UMAP depicting neighborhood (Nhood) groups identified by Milo and computed using the scVI representation. b, Beeswarm plot depicting differential abundance of neighborhoods (TiNK versus Ref-NK enriched). Colored neighborhoods are differentially abundant at a false recovery rate (FDR) of 0.1. c, Pie charts showing distribution of NK subsets across neighborhood groups annotated using our annotation model (Fig. 1). d, Expression of dominant TF regulons of NK cell differentiation across NK cell states (neighborhood groups). e, Expression of TF regulons uniquely expressed across cellular states. f, Graphic representation of cellular states. g,h, Volcano plots depicting DEGs between group 1 versus group 2 (g) and group 3 versus group 4/5/6 (h) cellular states. Differential expression analysis was performed using the findNhoodGroupMarkers method within the miloR package. Counts were aggregated per sample; groups were compared using edgeR and the adjusted P values were used for the plots. i, Scoring of pathway gene signatures in NK cell states. Func., function; homeo., homeostasis. j–n, Dot plots depicting selected genes belonging to stress response (j), immune suppression (k), metabolism (l), cytotoxicity (m) and chemokine/cytokine secretion (n). o, Pie charts depicting distribution of NK cell states in blood, tissues and tumors. Volcano plots: log(fold-change) cutoff at 0.5, P < 0.05. The scale represents regulon activity (d and e) or gene set activity (i) computed by AUCell.