| HA superfamily | EF fold | EF superfamily | EF family | PW | POOL-160 |
---|
SVM_P 5_1D1-3 | 23 | 24 | 25 | 24 | 27 | 24 |
SVM_P 24 | 21 | 24 | 24 | 23 | 30 | 23 |
SVM_P 24_1D1-3, 3D1-6 | 26 ◦ ∙ | 28 ◦ ∙ | 27 ◦ ∙ | 28 ◦ ∙ | 34 ◦ | 27 ◦ ∙ |
- Statistical comparisons of our best set of sequence-based features (SVM_P 5_1D1-3), the set of sequence-and structure-based features employed in Petrova and Wu [9] (SVM _P 24), and their combination with our additional set of structural neighborhood features (SVM_P 24_1D1-3, 3D1-6), excluding those coming from ligand information. Cross-validated F1measures (%) and results of a paired Wilcoxon test (α = 0.05) on the statistical significance of the performance differences are reported for all benchmark datasets employed in this study. A white circle indicates a statistically significant improvement of the classifier in the row over the sequence-based classifier (SVM_P 5_1D1-3), while a black bullet indicates a statistical significant improvement over the Petrova and Wu features (SVM_P 24).