Modular organization in the reductive evolution of protein-protein interaction networks

Table 1 Values of modularity for E. coli and Buchnera networks

Dataset	Modules and validation	Q_real	Q_rand	Q_norm (Q_real - Q_rand)
Newman algorithm
E. coli, Butland dataset	12 (5/10)	0.346	0.244	0.102
Buchnera, Butland dataset	7 (3/7)	0.259	0.232	0.027
Buchnera constrained, Butland dataset	7 (2/6)	0.182	0.168	0.014
E. coli, Arifuzzaman dataset	15 (8/13)	0.409	0.329	0.080
Buchnera, Arifuzzaman dataset	10 (4/9)	0.460	0.423	0.037
Buchnera constrained, Arifuzzaman dataset	12 (4/10)	0.274	0.265	0.009
E. coli, STRING	33 (32/32)	0.670	0.209	0.461
Buchnera, STRING	12 (11/11)	0.581	0.272	0.309
Buchnera constrained, STRING	14 (11/11)	0.493	0.210	0.283
Guimerá algorithm
E. coli, Butland dataset	10 (7/10)	0.357	0.248	0.109
Buchnera, Butland dataset	6 (3/5)	0.263	0.237	0.026
Buchnera constrained, Butland dataset	8 (2/7)	0.192	0.179	0.013
E. coli, Arifuzzaman dataset	12 (6/11)	0.413	0.332	0.081
Buchnera, Arifuzzaman dataset	8 (4/8)	0.461	0.432	0.029
Buchnera constrained, Arifuzzaman dataset	11 (2/8)	0.266	0.242	0.024
E. coli, STRING	19 (17/17)	0.669	0.211	0.458
Buchnera, STRING	11(10/10)	0.566	0.277	0.289
Buchnera constrained, STRING	9 (7/7)	0.489	0.231	0.258

Modularity is calculated using different algorithms as described in the text for the E. coli and Buchnera networks. The module validation is indicated between parentheses after the number of modules for each network and this provides information on the number of modules that are statistically significant with regards to the STRING data (see text for details). For instance, 5/10 means that five out of ten modules are significant in terms of STRING interactions. The number of modules validated is sometimes different to the total number of modules, since some modules are too small to be statistically assessed. When using STRING-derived networks, all modules can be validated since the same information was used to construct the network. The table also shows the modularity coefficient (Q) for real and randomized networks, and the normalized modularity coefficient, resulting from the subtraction of the modularity coefficients for real and random modules.

ISSN: 1474-760X