Extended Data Fig. 5: (related to Fig. 2). Cellular neighborhood assignment and (d-e) corresponding cell composition changes in T2D and ND islets.
From: Genetic risk converges on regulatory networks mediating early type 2 diabetes
(a) Heat map showing cellular neighborhoods (CNs) identified by CF-IDF method and representative islet image with cell and CN annotation overlay. See Methods for more details. Scale bar, 50 μm. (b-c) A k-means cellular neighborhood analysis (b) was applied in parallel, and results were highly concordant between the two methods (c; R2 = 0.9542, slope = 1.058, P < 0.0001; two-tailed linear regression). Scale bar (b), 50 μm. Bar graphs show mean + SEM with symbols representing individual donors. (d-e) Differential cell type enrichment (two-tailed t-test) in T2D vs. ND islets by CF-IDF (d) and k-means (e) analyses. ECs and pericytes were depleted in β CNs (CN1) of T2D islets, consistent with our findings of decreased proximity between β cells and ECs in T2D. Symbols indicate unadjusted P-values as follows: ◦ P < 0.2, • P < 0.1, *P < 0.05. (f-g) Correlation of cell compositions across CNs; these analyses ask whether cell type frequencies are correlated between CNs, i.e., if there was evidence for connectivity between spatially distinct regions. Panel (g) shows results from k-means method; see also Fig. 2l (CF-IDF method). All cellular neighborhood data is derived from CODEX imaging of tissue from n = 6 ND and n = 10 T2D independent donors.
