Figure 1. Integrated network analysis identified switch genes as putative key regulators of breast cancer transformation.

(a) Probability distribution of APCC for hubs (i.e. nodes with degree greater than 5 according⁶) identified in the correlation network built from the breast cancer expression dataset (blue solid line) and in its randomized counterpart obtained by shuffling the edges but preserving the degree of each node (grey dashed line). Differently from the randomized case, the true APCC distribution shows a clear trimodal pattern where peaks correspond to previously reported hub categories (such as party and date hubs) but also to the new category of hubs, the “fight-club hubs”. (b) Heat cartography map of nodes in the breast cancer correlation network. Dots in the plot, that correspond to nodes in the correlation network, are distributed across seven regions (R1 to R7) according to their clusterphobic coefficient K_π (x-axis), that is a measure of the propensity of a node to communicate with nodes in different modules, and to their within-module degree of communication z_g (y-axis). Dot color renders the APCC value in a red (positive APCC) to blue (negative APCC) scale. The black circle marks switch genes. (c) Node degree distribution in the breast cancer correlation network. The frequency distribution plot shows a clear power law behavior with most of the nodes marked by a small number of links in the network and a minority of nodes (hubs) marked by high links number. The same pattern is mirrored by coloring nodes based to decreasing degree on a red to purple scale (inset above the frequency distribution plot). (d) Analysis of network robustness. Scatter plots of average shortest path (y-axis) as a function of the fraction of removed nodes (x-axis) for targeted removal of different categories of hub nodes and for random node removal (see line labels in the plots). In the left panel, hubs are stratified according to the APCC-based classification of party, date, and fight-club hubs, while in the right panel hub nodes are stratified based on their topological characteristics in “switch” and “non-switch” hubs.