Table 1.
Cross-validation performance comparison of different MultiLoc architectures trained using the BaCelLo and the Höglund datasets
| Dataset | Method | Animals | Fungi | Plants | |||
| AVG | ACC | AVG | ACC | AVG | ACC | ||
| BaCelLo | |||||||
| MultiLoc | 77.3 (± 2.9) | 75.7 (± 3.1) | 78.4 (± 2.7) | 71.0 (± 2.6) | 71.4 (± 6.8) | 67.8 (± 3.8) | |
| + PhyloLoc | 80.1 (± 2.4) | 78.2 (± 2.9) | 80.0 (± 2.5) | 73.6 (± 0.9) | 78.6 (± 3.6) | 77.4 (± 1.9) | |
| + GOLoc | 84.0 (± 1.7) | 82.8 (± 2.0) | 81.1 (± 0.5) | 75.5 (± 1.1) | 80.9 (± 4.4) | 77.6 (± 3.5) | |
| MultiLoc2-LowRes | 86.1 (± 1.4) | 84.0 (± 1.7) | 82.8 (± 2.2) | 77.9 (± 0.5) | 81.9 (± 4.1) | 80.2 (± 3.5) | |
| Höglund | |||||||
| MultiLoc | 78.6 (± 1.2) | 76.4 (± 1.2) | 78.0 (± 1.3) | 76.6 (± 1.2) | 78.0 (± 1.8) | 76.4 (± 1.7) | |
| + PhyloLoc | 84.6 (± 0.7) | 84.0 (± 0.6) | 84.7 (± 1.4) | 84.4 (± 0.9) | 86.5 (± 1.5) | 84.3 (± 0.7) | |
| + GOLoc | 87.3 (± 1.8) | 86.7 (± 1.0) | 87.1 (± 0.9) | 86.9 (± 0.8) | 86.9 (± 1.4) | 86.3 (± 1.1) | |
| MultiLoc2-HighRes | 89.3 (± 1.4) | 88.6 (± 1.0) | 89.2 (± 1.1) | 88.9 (± 1.2) | 89.4 (± 0.8) | 88.7 (± 0.9) | |
This table compares the average sensitivities (AVGs) and overall accuracies (ACCs) of MultiLoc2-LowRes and MultiLoc2-HighRes with those of the original MultiLoc and the extended architecture based on PhyloLoc as well as GOLoc only. The AVGs and ACCs are given in percent. The standard deviations (in parentheses) refer to the differences of the AVGs and ACCs of the different cross-validation models.