Skip to main content

Table 4 The performance on the Gram-negative datasets

From: SigUNet: signal peptide recognition based on semantic segmentation

Method MCC (%) FPRTM (%) Precision (%) Recall (%) F1 measure (%)
The SignalP dataset
 Phobius 59.9 22.6 43.9 94.2 59.9
 PrediSi 30.6 69.0 19.7 86.5 32.1
 SignalP3.0-HMM 47.7 39.2 31.6 93.3 47.2
 SignalP3.0-NN 36.7 61.0 22.1 95.2 35.9
 PolyPhobius 60.7 21.4 45.0 94.2 60.9
 Philius 65.9 14.9 51.3 94.2 66.4
 SPOCTOPUS 64.7 17.0 50.8 92.3 65.5
 SignalP 4.0 84.8 1.5
 TOPCONS2 70.8 13.2 57.2 95.2 71.5
 DeepSig 81.2 1.7 88.9 76.9 82.5
 SigUNet 80.6 1.5 88.8 76.0 81.9
The SPDS17 dataset
 Phobius 69.5 18.0 56.4 95.7 71.0
 PrediSi 35.4 66.3 25.0 87.0 38.8
 SignalP3.0-HMM 65.4 21.3 51.2 95.7 66.7
 SignalP3.0-NN 49.1 44.9 33.8 95.7 50.0
 PolyPhobius 75.9 13.5 62.2 100.0 76.7
 Philius 88.7 2.2 84.6 95.7 89.8
 SPOCTOPUS 62.5 20.2 50.0 91.3 64.6
 SignalP 4.0 92.5 0.0 100.0 87.0 93.0
 TOPCONS2 85.9 5.6 76.7 100.0 86.8
 DeepSig 95.0 0.0 100.0 91.3 95.5
 SigUNet 97.6 1.1 95.8 100.0 97.9
  1. The best performance is highlighted in bold