Fig. 4

Performance comparison with eight existing protein complex prediction algorithms in terms of recall, precision, and F1-measure. Our method CPreditor3.0 achieves the highest F1-measure in five of the six experimental settings. (a) Results with Krogan as PPI dataset and MIPS as complex reference set, (b) Results with Krogan as PPI dataset and CYC2008 as complex reference set, (c) Results with Collins as PPI dataset and MIPS as complex reference set, (d) Results with Collins as PPI dataset and CYC2008 as complex reference set, (e) Results with WI-PHI as PPI dataset and MIPS as complex reference set, (f) Results with WI-PHI as PPI dataset and CYC2008 as complex reference set