Fig. 1

FINET parameter optimization and performance. a Frequency cutoff optimization. Frequency cutoff from 0.1 to 1.0 vs AUC, precision and normalized true positive calling (true positive callings at each cutoff/max(true positive callings at each cutoff)). This data resulted from FINET running on network1 at dream5 with following settings, m = 4, n = 500, alpha = 0.5 (see github software website for details). b Comparisons of precision of resampling m subgroups (frequency cutoff > 0.95). c, d The overall performance of FINET when m = 8 (c) and 12 (d). e Performance comparison between FINET, ARACNe-AP and C3NET. X-axis lab fo ARACNe-AP and C3NET represent p-value and alpha value, respectively, designed for significant threshold in ARACNe-AP and C3NET, while m in FINET as the number of sub-groups as shown above