Table 2 Performance of various modules of SVM developed by using Amino acid, dipeptide, split (2 and 4-part), PSSM and PSSM-PROSITE domain based input features
From: Analysis and prediction of cancerlectins using evolutionary and domain information
Methods | Threshold | Sensitivity | Specificity | Accuracy | MCC |
|---|---|---|---|---|---|
Amino acid Composition | -0.3 | 67.97 | 64.15 | 65.84 | 0.32 |
Dipeptide Composition | -0.3 | 67.27 | 62.84 | 64.84 | 0.30 |
Split based Composition (2-part) | -0.3 | 66.32 | 64.18 | 65.10 | 0.31 |
Split based Composition (4-part) | -0.5 | 65.12 | 66.85 | 66.09 | 0.32 |
Position-Specific Scoring Matrix | -0.2 | 67.92 | 68.57 | 68.34 | 0.36 |
PSSM with 14 PROSITE domains | -0.1 | 68.00 | 69.90 | 69.09 | 0.38 |