Skip to main content

Table 1 Performance of the inference algorithms on the DREAM4 networks (100 nodes)

From: Reconstruction of large-scale regulatory networks based on perturbation graphs and transitive reduction: improved methods and their evaluation

DREAM4 best performers     Score      
Team 395 (PG 1 + DR-FFL)     71.5889      
Team 296     71.2970      
Team 515 (PG 2 + TRANSWESD s,w,∞ )     64.7150      
Team 466     63.4060      
Team 549     63.1050      
Inference algorithm β γ α Score Edges TPs FPs FNs TP100
PG1 2.00 - - 70.3495 349.4 103.4 246.0 101.4 58.0
PG1 + DR-FFL - - - 71.5889 267.4 83.6 183.8 121.2 62.4
PG 1 + TRANSWESD u,w,∞ - - 0.95 73.0444 225.0 97.4 127.6 107.4 60.8
PG1 + TRANSWESDu,w,2 - - 1.50 79.6644 153.8 83.0 70.8 121.8 70.4
PG 1 + LTR u,u - - - 79.7428 261.2 92.4 168.8 112.4 72.2
PG1 + LTRu,w - - 0.15 79.1609 262.0 92.8 169.2 112.0 72.2
PG2 2.60 0.05 - 65.8012 398.2 98.0 300.2 106.8 58.2
PG2 + DR-FFL - - - 64.2614 372.8 94.4 278.4 110.4 58.0
PG2 + TRANSWESDu,w, - - 0.95 65.6504 256.6 86.0 170.6 118.8 64.4
PG2 + TRANSWESDs,w, - - 0.95 66.0970 260.8 86.8 174.0 118.0 66.6
PG2 + TRANSWESDu,w,2 - - 1.50 66.5562 224.0 81.0 143.0 123.8 64.2
PG 2 + TRANSWESD s,w,2 - - 1.50 68.1534 249.2 84.4 164.8 120.4 67.0
PG2 + LTRu,u - - - 65.4214 253.0 82.2 170.8 122.6 64.6
PG2 + LTRs,u - - - 67.7407 274.0 86.4 187.6 118.4 66.6
PG2 + LTRu,w - - 0.15 67.2567 271.4 85.6 185.8 119.2 66.0
PG 2 + LTR s,w - - 0.15 68.5959 288.2 88.4 199.8 116.4 67.2
PG new 2.00 0.00 - 81.7594 250.2 99.6 150.6 105.2 66.6
PGnew + DR-FFL - - - 80.3085 179.8 82.0 97.8 122.8 66.2
PGnew + TRANSWESDu,w, - - 0.95 85.3288 179.2 90.2 89.0 114.6 72.0
PGnew + TRANSWESDs,w, - - 0.95 85.7898 183.0 92.0 91.0 112.8 72.8
PGnew + TRANSWESDu,w,2 - - 1.50 88.0570 147.6 86.0 61.6 118.8 72.6
PG new + TRANSWESD s,w,2 - - 1.50 88.5728 150.4 87.8 62.6 117.0 72.8
PGnew + LTRu,u - - - 88.2217 166.8 91.8 75.0 113.0 74.2
PGnew + LTRs,u - - - 88.6350 169.4 93.4 76.0 111.4 75.2
PGnew + LTRu,w - - 0.15 88.5203 168.4 92.8 75.6 112.0 75.0
PG new + LTR s,w - - 0.15 88.8005 169.8 93.8 76.0 111.0 75.8
  1. The table summarizes the performance (overall score, true positives (TPs), false positives (FPs) and false negatives (FNs)) of the different PG generation and TR algorithms when applied to the DREAM4 networks together with the displayed (optimal) parameters. The last column TP100 shows the average number of TPs within the first 100 top-ranked (reconstructed) edges. As a comparison, the scores of the 5 best-performing algorithms within the challenge are shown. PG2 denotes PG2 computed with a minor bug in the original implementation.