Simulation modelling
To test for non-random patterns of weighted degree between each sex, we applied a simulated network annealing approach from the Ergm R package (Handcock et al, 2018). To generate each simulated network, we fitted a variety of probability distributions to edge weight and degree of both treatment and stable regions, then used AIC to select the best fitting target distribution. Edge density, network size and the number of isolated nodes were fixed based on each observed network. We added sex to each simulated node attribute drawing from a Bernoulli distribution (probability= 0.5). Using these network characteristics, we generated 1000 ‘null’ networks and compared the homophily weighted degree distribution of each sex (i.e., the average weighted degree for each individual based on putative male-male or female-to-female transmission events) of the null networks to the observed and calculated a bootstrap p-value.