Skip to main content

Table 6 The performance of data augmentation on the SignalP dataset

From: SigUNet: signal peptide recognition based on semantic segmentation

Comp Eukaryotes Gram-positive Gram-negative
Train MCC (%) FPRTM (%) MCC (%) FPRTM (%) MCC (%) FPRTM (%)
SigUNet
 As compa 90.2 4.0 76.1 5.1 80.6 1.5
 All organismsb 89.9 3.2 80.9 3.1 82.1 3.6
 Bacteriac 79.3 1.9 83.5 0.3
SigUNet-light
 As comp 89.4 4.3 77.7 5.1 82.9 1.9
 All organisms 88.9 3.9 82.5 3.1 81.4 3.5
 Bacteria 80.2 1.9 83.9 2.7
  1. aThe model is trained using the same organism as the comparison dataset. bThe model is trained using all organisms. cThe model is trained using all of the bacteria data. The best performance is highlighted in bold