Skip to main content

Table 2 The performance on the Eukaryotes datasets

From: SigUNet: signal peptide recognition based on semantic segmentation

Method MCC (%) FPRTM (%) Precision (%) Recall (%) F1 measure (%)
The SignalP dataset
 Phobius 81.1 15.3 77.6 95.2 85.5
 PrediSi 56.1 52.6 52.0 91.3 66.3
 SignalP3.0-HMM 75.9 23.5 69.5 97.4 81.1
 SignalP3.0-NN 56.2 64.1 48.4 98.8 65.0
 PolyPhobius 80.6 12.5 79.5 91.9 85.2
 Philius 80.4 13.4 77.8 93.7 85.0
 SPOCTOPUS 80.1 14.0 79.0 91.7 84.9
 SignalP 4.0 87.4 6.1
 TOPCONS2 84.6 9.6 83.6 93.6 88.3
 DeepSig 87.2 4.2 92.5 87.8 90.1
 SigUNet 90.2 4.0 93.0 92.1 92.5
The SPDS17 dataset
 Phobius 65.8 9.6 47.8 95.7 63.8
 PrediSi 38.5 43.3 20.7 89.1 33.6
 SignalP3.0-HMM 51.6 22.3 31.2 95.7 47.1
 SignalP3.0-NN 36.0 59.1 17.5 95.7 29.5
 PolyPhobius 72.0 8.0 56.4 95.7 71.0
 Philius 62.3 6.5 44.3 93.5 60.1
 SPOCTOPUS 54.0 16.4 37.9 84.8 52.3
 SignalP 4.0 81.9 4.0 75.0 91.3 82.3
 TOPCONS2 73.9 5.6 60.6 93.5 73.5
 DeepSig 86.1 2.5 82.4 91.3 86.6
 SigUNet 89.6 1.2 91.1 89.1 90.1
  1. The performances of Phoibus, PrediSi and SignalP 3.0 are obtained from their online services (Phobius: http://phobius.sbc.su.se/; PrediSi: http://www.predisi.de/predisi/; SignalP 3.0: http://www.cbs.dtu.dk/services/SignalP-3.0/) [11, 23, 24]. The performances of PolyPhobius, Philius, SPOCTOPUS and TOPCONS2 are obtained from the TOPCONS2 software (https://github.com/ElofssonLab/TOPCONS2) [25,26,27,28]. The performance of SignalP 4.0 on the SignalP dataset is obtained from the original paper [12] and the performance on the SPDS17 dataset is obtained from its online service (http://www.cbs.dtu.dk/services/SignalP-4.0/). The performance of DeepSig on the SignalP dataset is obtained by reproducing the algorithm and the performance on the SPDS17 dataset is obtained using the source code (https://github.com/BolognaBiocomp/deepsig). For each dataset, the best performance is highlighted in bold.