Article Extended Data Fig. 2 | Characterizing the melanoma TDLN microenvironment. ( a ) UMAP projections of scaled expression levels of the indicated marker proteins across all cells identified in the combined IMC experiments. ( b ) Sankey plot showing relationships between two different layers of annotation used in the study: level 1 (left) and level 2 (right). ( c ) Changes in cellular composition depicted as log 2 fold changes in TDLNs versus cLNs of mean number of cells/mm² tissue. Only those changes with an adjusted P < 0.05 are shown. ( d ) Composition of ROIs, separated by location (IT = intratumoral, PT = peritumoral, PC = paracortex, C = cortex) in different disease outcome groups, displayed as proportions of total cells per group. ( e ) Number of macrophages (Mac) and DCs per mm 2 in the paracortex of the TDLNs of stage III disease-free (DF) and recurrence (R) patients with melanoma, visualized per ROI (R = 24 & DF = 30 ROIs from 5 or 6 patients, respectively; 4–5 ROIs per patient). ( f ) DC/Mac ratio in cLNs and the TDLN paracortex of patients with stage II or III melanoma, visualized per ROI (R = 24 & DF = 30 ROIs from 5 or 6 patients respectively; 4–5 ROIs per patient). ( g ) Comparison of Ilastik and Mesmer cell segmentation pipelines revealed no significant ( P > 0.05) differences in cell-type abundance across ROI (n = 352 ROIs). ( h ) Volcano plots show similar changes in abundance (cells/mm 2 ; P < 0.05) of cell types for both segmentation methods when comparing stage III R versus DF patients. Only a small subset of cells showed borderline method-specific significant changes. P values were calculated using an unpaired two-sided t-test ( e – h ). Data are presented as mean ± s.e.m. ( e – g ). * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 (Exact P values in Supplementary Data).