Phylogentic topology similarity measure results. (a) Topological matching for LC interaction graph: Boxplots for the distribution of matching topologies found for each sampled graph ensemble. The red line shows the result for the LC graph (also in (b) and (c)). The [complex] constrained graphs are the only ensembles that present fewer matching topologies than are found in the empirical data. (b) Mismatch score using LC interaction graph: Node shuffle and random graph ensembles exhibit higher scores than found in the empirical data, whilst network shuffle ensembles under any of the tested constraints exhibit a lower average score than is found in the empirical graph. (c) Topological similarity for LC interaction graph: The empirical data have similar similarity values as those found for the network shuffle graphs, whilst the similarity is significantly higher for each of these than is found in graphs sampled from the random graph ensemble. For each plot, the results are shown for the simple topological ensembles and also those graph ensembles which are constrained using biological traits. The three different plots highlight the sensitivity of the results to the metric used, as well as the different interpretations that may be found given different null graph models used for comparison.