| SVM model | Scoring scheme |
γ
|
C
| 5-fold ACC |
---|
 | SVMT 1 | OvO | 16.0 | 32.0 | 99.01% |
Level 2: Rhodopsin | SVMT 7 | OvO | 2-8 | 2048.0 | 99.47% |
 | SVM
FL
| OvO | 256.0 | 32.0 | 98.65% |
 | SVMT 1 | OvO | 256.0 | 32.0 | 96.44% |
Level 3: Amine | SVMT 7 | OvO | 32.0 | 24.5 | 95.0% |
 | SVM
FL
| OvO | 256.0 | 32.0 | 94.77% |
- RBF grid-search parameters used to train SVM models, along with the corresponding 5-fold cross-validation accuracy (ACC) for the training set. The best performing model for level 2 (subfamily) classification is SVMT 7, while the best performing model for level 3 (amine) classification is SVMT 1. Both classifiers exclusively employ transmembrane sequence data, and outperform the classifiers trained with full-length sequence data.