ROC curves for using mean clustering coefficient to pick out functionally homogeneous communities in a) the A network and b) the P network. The Receiver Operating Characteristic (ROC) curve for using mean clustering coefficient as a predictor of functional homogeneity under the GO measure (solid green curve), MIPS measure (dot-dashed blue curve) and correlated growth measure (dashed red curve). We plot the false positive rate (FPR) versus the true positive rate (TPR). A random classifier would give the solid black line. For the A network under the GO measure, a true positive rate of 70% is achievable with a false positive rate of 30%. For both networks, the best predictive ability is achieved for the GO measure, and the worst for the MIPS measure (see Table 4 for areas under the curves (AUCs).). The AUCs for the P network are in general lower than those for the A network (see Table 4).