Prediction performance of SVM models used in SSPred
| The SVM Models were developed using SVM-light package utilizing fixed length vector derived from amino-acid, dipeptide composition, physico-chemical properties, hyrbid-I and hybrid-II based input features. The performance of the trained modules was measured in terms of Mathew Correlation Coefficient (MCC) and was evaluated using 5-fold cross validation and validation test. |
| Training dataset (5-fold cross validation) |
| Class | SVM Model |
| amino-acid | dipeptide | physico-chem | hybrid-I | hybrid-II |
| Secretion System | 0.73 | 0.71 | 0.69 | 0.75 | 0.80 |
| Type-I | 0.85 | 0.90 | 0.83 | 0.87 | 0.95 |
| Type-II | 0.52 | 0.55 | 0.46 | 0.60 | 0.71 |
| Type-III | 0.70 | 0.76 | 0.68 | 0.78 | 0.85 |
| Type-IV | 0.65 | 0.70 | 0.63 | 0.67 | 0.72 |
| Sec | 0.65 | 0.70 | 0.63 | 0.68 | 0.76 |
|
| Test dataset (Validation test) |
| Class | SVM Model |
| amino-acid | dipeptide | physico-chem | hybrid-I | hybrid-II |
| Secretion System | 0.69 | 0.71 | 0.67 | 0.74 | 0.81 |
| Type-I | 0.82 | 0.88 | 0.78 | 0.84 | 0.92 |
| Type-II | 0.49 | 0.55 | 0.41 | 0.57 | 0.72 |
| Type-III | 0.74 | 0.75 | 0.66 | 0.78 | 0.88 |
| Type-IV | 0.64 | 0.57 | 0.56 | 0.61 | 0.64 |
| Sec | 0.60 | 0.68 | 0.57 | 0.66 | 0.73 |
|
| Independent dataset (Validation test) |
| Class | SVM Model |
| amino-acid | dipeptide | physico-chem | hybrid-I | hybrid-II |
| Secretion System | 0.78 | 0.78 | 0.81 | 0.85 | 0.92 |
| Type-I | 0.56 | 0.49 | 0.42 | 0.66 | 0.86 |
| Type-II | 0.16 | 0.21 | 0.07 | 0.09 | 0.53 |
| Type-III | 0.53 | 0.45 | 0.34 | 0.39 | 0.64 |
| Type-IV | 0.60 | 0.60 | 0.67 | 0.67 | 0.78 |
| Sec | 0.38 | 0.42 | 0.44 | 0.34 | 0.76 |
|
