Fig. 1

Evaluation of EnDED: intersection combination and individual methods on simulated networks. Using 1000 simulated networks, and 1000 simulated networks incorporating noise, we evaluated EnDED’s performance. A The evaluation measurements true positive rate (TPR), true negative rate (TNR), accuracy (ACC), and positive predictive value (PPV) for each individual method, i.e., sign pattern (SP), overlap (OL), interaction information (II), and data processing inequality (DPI), as well as the intersection combination (COMBI). SP and OL perform best according to TPR and ACC, while the intersection combination performs best according to TNR. All methods performed well according to PPV. The intersection combination, DPI and II performed better on noisy data according to TNR because less edges were removed along with less true interactions. B The ROC curve for each environmentally driven edge detection method as well as their intersection combination