Overview of the performance of various module preservation statistics in our simulation studies.

*Value not available since the pathway membership is determined beforehand and the same in both data sets.§Since no thresholds can be defined for the statistic , the top grade of 4 was assigned if the perfectly distinguished preserved from non-preserved modules, i.e. if the of each preserved module was smaller than that of all non-preserved modules. For simulated scenario 3 (Perm.), a grade is not available (NA) since none of the modules were simulated to be preserved.Evaluating preservation statistics in different simulation scenarios. Best and worst performance correspond to grade 4 and 1, respectively. The columns report the name, type, and input of individual preservation statistics, and their performance our 7 simulation studies summarized in Figure 9: 1. weak module preservation, 2. strong module preservation, 3. permuted module membership, 4. half-permuted module membership, 5. in-module permutation, 6. small pathway, and 7. large pathway. We graded each statistic as follows: Statistics that accurately distinguish preserved and non-preserved modules within the thresholds of Z≲2 for non-preserved modules and Z≳10 for preserved modules get grade 4; statistics that accurately distinguish preserved and non-preserved modules but whose values may lie on the wrong side of the thresholds are graded 3; statistics that distinguish preserved and non-preserved modules with high accuracy (allowing 20%, that is 1 of 5, mis-classifications) are graded 2, and statistics that perform worse are graded 1. The grading of is explained in . We only report selected statistics defined for correlation networks.