Table 2.
Ranking of methods with respect to the maximum F1 score of the threshold measure curves.
| F1 BPO | Rank BPO | F1 MFO | Rank MFO | |
|---|---|---|---|---|
| Presented @ CAFA | ||||
| GOtcha | 0.29 | 13 | 0.47 | 4 |
| BLAST | 0.21 | - | 0.34 | - |
| Priors | 0.27 | 15 | 0.41 | 12 |
| StudentA | 0.32 | 8 | 0.40 | 13 |
| StudentB | 0.15 | - | 0.20 | - |
| StudentC | 0.28 | 14 | 0.36 | - |
| Best@CAFA | 0.37 | 1 | 0.49 | 1 |
| Post-CAFA | ||||
| Priors' | 0.20 | - | 0.29 | - |
| Student A' | 0.33 | 8 | 0.43 | 10 |
| Student B' | 0.36 | 3 | 0.45 | 7 |
| Student C' | 0.34 | 6 | 0.48 | 3 |
| MetaStudent' | 0.36 | 3 | 0.48 | 3 |
This table shows the maximum F1 score (Fmax) of each threshold measure curve in Figure 5 and its rank in the list of competing methods which was shown at CAFA. This list actually consists of 36 predictors, but only the scores and ranks of the top 15 performers have been released. Classifiers which are actually part of this list are kept in bold. Ranks of other methods are hypothetical, either because calculated after CAFA or because discarded by the CAFA organizers. They considered only one method per participating group and we chose method A. Results for StudentB were compiled with the bug (Methods).