Table 2. Comparison of the PupPred with existing methods.
Method | Threshold | Ac(%) | Sn(%) | Sp(%) | MCC |
GPS-PUP | High | 85.44 | 33.07 | 90.18 | 0.1991 |
Medium | 82.51 | 44.88 | 85.91 | 0.2279 | |
Low | 78.85 | 63.78 | 80.21 | 0.2864 | |
iPUP | 85.24 | 70.87 | 86.35 | 0.3860 | |
81.92 | 78.74 | 82.16 | 0.3777 | ||
78.25 | 85.04 | 77.73 | 0.3654 | ||
PupPred1 | 85.07 | 30.71 | 89.26 | 0.1574 | |
83.15 | 42.52 | 86.29 | 0.2037 | ||
77.58 | 54.33 | 79.37 | 0.2063 | ||
PupPred | 92.97 | 97.27 | 92.66 | 0.6629 | |
88.29 | 98.18 | 87.57 | 0.5582 | ||
79.90 | 100.00 | 78.44 | 0.4449 |
The thresholds of High, Medium and Low for GPS-PUP represent that the score are greater than 2.738, 2.452 and 2.111, respectively; the thresholds of High, Medium and Low for iPUP represent that the score are greater than 0.1167, 0.1044 and 0.0963, respectively; the thresholds of High, Medium and Low for PupPred represent that the SVM probability are greater than 0.8, 0.7 and 0.5.