The global networks created by ligand-receptor mapping are highly multimodal, with ~0.25 × 106 edges divided between ~1 × 103 specific ligand-receptor pairwise mechanisms (modalities). For many mechanisms, edges were promiscuous, connecting most nodes in the graph together combinatorically due to similar levels of gene expression in many cell types. For the global network comparisons in Fig. 3, we were predominantly interested in outlier connections where ligand and receptor expression were differentially expressed above the mean of the global community, and therefore, we limited our dataset to only those edges where the cluster average z score was positive for both ligand and receptor. Weights of edges were defined as the sum of the z scores for ligand and receptor (see Supplementary Text). Global node-node affinity, when used, was defined as the sum of the weights of all edges connecting pairs of nodes. Nodal degree, hub, authority, and eigenvector centrality were calculated using the “igraph” package in R (54). For Fig. 3C, hub scores were calculated using this global node-node global affinity. For Fig. 3 (D and E), all edges above the percentage threshold were incorporated into calculations of degree and centrality. Thresholding graphs were created by iteratively subsetting the edge list for each species to include only those edges above a given percentage expression and then recalculating degree distributions, nodal centrality rankings, and Spearman correlations for each threshold. Spearman correlations were calculated by aligning node order across species and comparing centrality metrics using the human as a reference. Network graphs were created using the igraph package in R.