How biologically relevant are interaction-based modules in protein networks?

Table 1 Global and follow-up analysis of the network modular organizations

For every functional network of size n, we applied the network clustering algorithm with a given number of branches in the clustering tree, . These -values were chosen to be among those with significantly high average maximal overlap, that is, overlap equal to or greater than 0.8, low overlap ratios, and meso-scale average module size, that is, ~5-25. The outcome of this algorithm is a modular organization with M modules. For the follow-up analysis of both full and core components of the modules, third and fourth column groups, the following quantities are shown: ξ, the overall statistic, P, statistical significance of global test, P_m^†, number of modules whose branch length in the similarity dendrogram (see text for details) is bigger than 0.1 in similarity units and P_m, number of modules whose within-similarity is statistically significant (P < 0.05) in the modular test. All P-values were obtained by means of an approximate permutation test with 10,000 randomizations and the use of binary phylogenetic profiles with a threshold of E_th = 1e^-6 in the BLAST E-value [35].

ISSN: 1474-760X