Figure S2 Comparison of graph permutation and gene permutation on two representative simulation data sets. (a) Simulation data set (r = 0.5, p = 50%). (b) Simulation data set (r = 0.3, p = 70%). Scoring function "ratio" is used in both cases. Both permutation methods can reach good sensitivity and specificity. However, graph permutation reaches slightly better specificity in most cases.