Figure 4

A. Comparison of topological measures in identification of important proteins. Networks were inferred as described in the text, varying parameters (see Additional file 3: Table S2), and topological measures calculated (X axis). The enrichment of known pathogen interactors in the top 20% of proteins ranked by each measure was calculated and is shown as the percentage enrichment in the group divided by the enrichment in background. The boxes represent the 25^th and 75^th percentiles, the bold bar represents the mean, and the dashed bars are at 1.5 times the interquartile range. The enrichment given by differential abundance is shown as a comparison. This figure shows that degree and betweeness perform much better than other, more sophisticated, measures of importance for these networks. B. Comparison of the enrichment of topological sub-types. For the cell culture co-abundance network integrated with PPIs we assessed statistical enrichment in pathogen targets for the following sub-types of topology: NH-NH, non-hub non-bottleneck nodes; HB, hubs; BN, bottlenecks; H-NB, hub non-bottlenecks; B-NH, bottleneck non-hubs; BH, bottleneck hubs. These sub-types were determined as the overlap of the top 20% of nodes ranked by degree (hubs) or betweenness (bottlenecks). The analysis shows that betweenness is the primary driver of importance in these networks similar to observations in directed regulatory networks.