Skip to main content

Advertisement

Table 4 Topological metrics tested and AUCs

From: The function of communities in protein interaction networks at multiple scales

  A P
Network topology measure G C M G C M
Mean degree 0.6476 0.6476 0.6142 0.5130 0.5373 0.5387
Degree assortativity coefficient [58] 0.6913 0.6913 0.6277 0.4799 0.5517 0.5181
Clustering coefficient[59] 0.7186 0.7186 0.6613 0.5521 0.5829 0.5725
Global mean Soffer clustering coefficient [60] 0.4857 0.4857 0.4819 0.3915 0.4735 0.4461
Local mean Soffer clustering coefficient [60] 0.4784 0.4784 0.4662 0.3892 0.4654 0.4540
Mean geodesic node betweenness centrality [61] 0.4600 0.4600 0.4973 0.5045 0.5094 0.4959
Mean closeness centrality [61] 0.5275 0.5275 0.5524 0.4877 0.4919 0.4815
Mean eigenvector centrality [61] 0.5601 0.5601 0.5722 0.5312 0.5551 0.5246
Mean information centrality [61] 0.5191 0.5191 0.5429 0.5253 0.5456 0.5170
Mean geodesic distance [59] 0.3839 0.3839 0.3717 0.4274 0.4945 0.5066
Diameter [61] 0.4457 0.4457 0.4042 0.4366 0.5004 0.5079
Mean harmonic geodesic distance [59] 0.4088 0.4088 0.4042 0.5024 0.4834 0.4995
Energy [59] 0.5237 0.5237 0.4982 0.4568 0.4976 0.5114
Entropy [59] 0.5655 0.5655 0.5327 0.5077 0.5127 0.5280
Off-diagonal complexity [62] 0.5941 0.5941 0.5457 0.5081 0.5054 0.5237
Cyclomatic number [62] 0.6331 0.6331 0.5733 0.5173 0.5300 0.5425
Connectivity [62] 0.6437 0.6437 0.5766 0.5245 0.5334 0.5468
Number of spanning trees [62] 0.4525 0.4525 0.4531 0.4451 0.4516 0.4491
Medium articulation [62] 0.5659 0.5659 0.4463 0.5295 0.5070 0.5592
Efficiency complexity [62] 0.5316 0.5316 0.5343 0.4911 0.4945 0.4982
Graph index complexity [62] 0.6564 0.6564 0.6492 0.5211 0.5469 0.5250
Density 0.6541 0.6541 0.6553 0.5277 0.5676 0.5235
Efficiency [63] 0.5790 0.5790 0.5896 0.4964 0.5071 0.4865
Fraction of articulation vertices [64] 0.5065 0.5065 0.5028 0.5216 0.5062 0.5091
Largest eigenvalue 0.6054 0.6054 0.5663 0.4941 0.5041 0.5185
Rich club coefficient [65] 0.5428 0.5428 0.5896 0.4988 0.5209 0.4868
  1. The network topology measures tested and their associated AUCs. We report the results for using each of these as a predictor for functional homogeneity as judged under the three measures of functional similarity (GO, G, correlated growth rates, C, and MIPS, M) for both the A and P networks. The AUCs are given as the average performance over the range 0 ≤ log(λ) ≤ 3. The clustering coefficient (definition given in the text, equation 9) is the best predictor in all cases. (The topological properties were computed from code developed by Gabriel Villar.)