Figure 3.

Small-world network structure is maintained in post-mortem networks in disease states – example from human amygdala. (A) Connectivity of hubs (top 5% connectivity nodes) and equal number of provincial (non-hub) nodes for examination of targeted differential connectivity in disease. This example selection of two different types of genes will be used to illustrate that disease does not target hub connections. (B) For this selection of provincial and hub nodes, we compute the relative fraction of altered links (both created and destroyed) between control and depressed networks. For comparison the same rewiring statistic is calculated for disease-permutated data (pseudo network comparisons with no disease effect). These permutations establish the mean and expected confidence bounds on a real effect. While provincial nodes are generically more likely to show differential connectivity compared to hub nodes, this is not a disease effect, but rather due to the greater statistical stability of hub nodes with a large base of connections, since the “real” network connectivity changes are within the expected bounds of variability. (C) To further quantify panel 3B, the p-values for probability of greater than expected differential connectivity are plotted, showing that connectivity changes in depression are not greater than expected by chance for both provincial and hub nodes.