Community Landscapes: An Integrative Approach to Determine Overlapping Network Module Hierarchy, Identify Key Nodes and Predict Network Dynamics
Overlapping modules of the yeast protein-protein interaction network of Ekman et al.  were identified using the LinkLand influence function calculation method with the TotalHill module membership assignment method using the modular structure of the lowest level of hierarchy. During the post-processing of the module assignment, we merged the modules with ProportionalHill module membership assignment-based correlation higher than 0.9 (see Section VI. in the Electronic Supplementary Material S1, we received similar results without this merging process; data not shown). Panel A: 3D view of the yeast protein-protein interaction network. The underlying 2D network layout was set by the Kamada-Kawai algorithm. The vertical positions reflect the community landscape values of the nodes on a linear scale. Nodes were colored as the module of their maximum membership. Panel B: centrality and bridgeness of yeast date- and party-hubs. Hubs having more than 8 neighbors and non-hubs with less neighbors were positioned on the scattergram according to their ModuLand centrality (x-axis, the height of the community landscape) and ModuLand bridgeness (y-axis) as defined in Section V.6.d. in the Electronic Supplementary Material S1. Date- and party-hubs are marked with red circles and blue triangles, respectively, while non-hub proteins are represented by gray crosses. The inset shows a double logarithmic plot of hubs with large centrality.