Figure 3

Validation of two-way module-finding-hierarchy-building strategy. Pyramabs was validated using 640-node hierarchically nested random networks. (A) The pre-specified nested structures of the random network. (B) An example of a 640-node hierarchically nested random network. Thirty random networks were generated for validation using various values for parameter ρ, which controls the probability that an edge between nodes i and j exists. The smaller the value of ρ, the more cohesive the levels (see Supporting Information of [4]). (C) The measured accuracy based on the mutual information between the predicted and the real partitions. The accuracy averaged over 30 random networks is provided in the histogram, and shows that Pyramabs was comparable to box clustering, and performed better than the standard module-finding algorithm based on edge betweenness and a modularity measure of the GN algorithm [12, 13]. (D) The performance of Pyramabs depending on density. (E) The performance of Pyramabs depending on z-score threshold values.