Table 4.
Comparison of accuracy when predicting the structural classes using all features, each feature individually, and when excluding one features at the time.
| Features | Accuracy when predicting with one feature | |||||
| all-α | all-β | α/β | α+β | overall | ||
| All features included | 92.8 | 80.6 | 74.3 | 71.4 | 80.1 | |
| using only one feature | PSIPRED-NCountH6 | 89.6 | 58.7 | 32.9 | 46.7 | 58.4 |
| PSIPRED-NCountH8 | 81.9 | 78.3 | 53.8 | 58.7 | 69.0 | |
| PSIPRED-CMVH1 | 76.3 | 74.5 | 55.8 | 48.8 | 64.3 | |
| PSIPRED-NAvgSegH | 49.9 | 83.3 | 0.0 | 47.8 | 47.9 | |
| PSIPRED-NCountE5 | 85.8 | 59.1 | 50.9 | 47.4 | 61.4 | |
| PSIPRED-CVE | 88.9 | 71.3 | 60.7 | 51.0 | 68.4 | |
| PSIPRED-MaxSegE | 83.1 | 48.8 | 0.0 | 67.1 | 52.6 | |
| PSIPRED-NAvgSegE | 79.2 | 33.9 | 3.2 | 42.4 | 41.8 | |
| CVL---G | 73.8 | 0.0 | 54.3 | 7.7 | 32.8 | |
| excluding the listed feature | PSIPRED-NCountH6 | 92.1 | 79.5 | 71.7 | 70.8 | 78.9 |
| PSIPRED-NCountH8 | 93.0 | 79.5 | 73.1 | 70.3 | 79.3 | |
| PSIPRED-CMVH1 | 92.5 | 80.8 | 72.5 | 71.0 | 79.6 | |
| PSIPRED-NAvgSegH | 92.5 | 81.0 | 71.4 | 68.7 | 78.8 | |
| PSIPRED-NCountE5 | 90.7 | 80.1 | 73.4 | 71.4 | 79.3 | |
| PSIPRED-CVE | 91.9 | 80.6 | 72.5 | 71.4 | 79.5 | |
| PSIPRED-MaxSegE | 92.8 | 79.7 | 73.1 | 69.8 | 79.2 | |
| PSIPRED-NAvgSegE | 92.3 | 80.6 | 71.1 | 69.2 | 78.7 | |
| CVL---G | 92.8 | 80.6 | 73.4 | 68.9 | 79.3 | |
Bold font shows the top two highest accuracies when using individual features, and features selected for further analysis.