The Impact of Past Epidemics on Future Disease Dynamics: Supplementary Information

In Figure 1, we compare additional structural properties of the residual network to those of the original contact network. In particular, we consider transitivity, or the propensity of triangles in the network, and the degree correlation coefficient. To consider the structural impact of the epidemic process in the most extreme situations, we focus on two sets of parameter values: (a) α = 0, T1 ∈ [0, 1], (b) T1 = 1, α ∈ [0, 1] for two sets of average degrees (3, 10) in random networks distributed with Poisson, exponential and scale-free degree distributions. The residual network is attained by performing 500 Monte Carlo SIR percolation simulations on each original random network. As shown below, for all parameter values, absolute values of clustering and degree correlations in the residual network remain small (or decrease) compared to the original network, and thus the assumptions of bond percolation still apply. The scale-free distributed network has large transitivity and disassortativity originally, but absolute values of both measures are reduced due to the infection process.