Skip to main content


Fig. 3 | BMC Systems Biology

Fig. 3

From: Identification of regulatory modules in genome scale transcription regulatory networks

Fig. 3

Evaluation of the different module-finding methods using real networks. We used different similarity indices for CoReg, CoReg_inv: CoReg + inverse log weighted similarity index; CoReg_jaccard: CoReg + jaccard similarity index; CoReg_geometric: CoReg + geometry index. CoReg was also compared to other three clustering algorithms, namely, Label Propagation (LP), Edge Betweenness (EB), Walk Trap (WT). We performed the evaluation on A. thalina, E. coli and H. sapiens network, respectively (From left to right, species was indicated on the top of the figure). a, b, c The ROC curve for co-regulators pairs based on the ranking result from CoReg and WT. d, e, f Rewiring recall score for all the methods. We calculated rewiring recall score under rewiring probability from 0 to 0.5. Each data point is the average score of five runs. Error bar was added to show the standard error. For the human network, EB algorithm was not tested because computation cannot be finished within a reasonable amount of time on large-scale network such as human network

Back to article page