Table 2. Results of the application of network inference algorithms on the simulated data set
ARACNEBANJONIRClusteringRandom
PPVSePPVSePPVSePPVSePPV
Global (steady‐state)
 10 × 100.37u0.40u0.41u0.49u0.34u0.71u0.40u0.38u0.36u
0.25d0.17d0.18d0.45d0.20d
0.16s0.05s0.09s0.22s0.10s
 10 × 1000.37u0.44u0.96u0.11u0.36u0.70u0.36u0.36u0.36u
0.79d0.05d0.20d0.46d0.20d
0.84s0.05s0.09s0.21s0.10s
 100 × 100.19u0.11u0.19u0.04u0.18u0.09u0.19u0.11u0.19u
0.10d0.02d0.10d0.05d0.10d
0.06s0.00s0.05s0.02s0.05s
 100 × 1000.19u0.17u0.70u0.00u0.19u0.19u0.19u0.11u0.19u
0.47d0.00d0.10d0.10d0.10d
0.71s0.00s0.05s0.05s0.05s
 100 × 10000.19u0.26u0.99u0.05u0.20u0.19u0.19u0.11u0.19u
0.68d0.03d0.10d0.09d0.10d
0.68s0.03s0.05s0.05s0.05s
 1000 × 10000.02u0.10u0.02u0.01u0.02u
Local (steady‐state)
 10 × 100.53u0.61u0.41u0.50u0.63u0.96u0.39u0.38u0.36u
0.25d0.18d0.57d0.93d0.20d
0.15s0.05s0.57s0.93s0.10s
 100 × 1000.56u0.28u0.71u0.00u0.97u0.87u0.29u0.18u0.19u
0.42d0.00d0.96d0.86d0.10d
0.60s0.00s0.96s0.86s0.05s
 1000 × 10000.66u0.65u0.20u0.10u0.02u
Dynamic (time‐series)
 10 × 1000.39u0.36u0.35u0.35u0.33u0.36u
0.22d0.21d0.20d
0.00s0.00s0.10s
 10 × 1000.35u0.43u0.36u0.29u0.35u0.33u0.36u
0.21d0.16d0.20d
0.25s0.00s0.10s
 100 × 100.19u0.10u0.18u0.08u0.19u0.12u0.19u
0.10d0.04d0.10d
0.06s0.00s0.05s
 100 × 1000.19u0.15u0.19u0.05u0.19u0.11u0.19u
0.10d0.02d0.10d
0.04s0.00s0.05s
 100 × 10000.19u0.19u0.19u0.04u0.19u0.11u0.19u
0.10d0.02d0.10d
0.05s0.00s0.05s
 1000 × 10000.02u0.10u0.02u0.01u0.02u
  • Abbreviations: PPV: positive predicted value; Se: sensitivity.

  • In bold are the algorithms that perform significantly better than random, using as a random model a Binomial distribution.