Log in to download the datasets
| Models & Date | Swiss-Prot | PDB | KEGG |
|---|---|---|---|
| 04-01-2026 | SwissProt_04-01-2026 1 | PDB_04-01-2026 2 | KEGG_04-01-2026 3 |
| 01-03-2026 | SwissProt_01-03-2026 9 | PDB_01-03-2026 3 | KEGG_01-03-2026 4 |
| 10-01-2025 | SwissProt_10-01-2025 6 | PDB_10-01-2025 2 | KEGG_10-01-2025 4 |
| 07-02-2025 | SwissProt_07-02-2025 5 | PDB_07-02-2025 2 | KEGG_07-02-2025 2 |
| 04-01-2025 | SwissProt_04-01-2025 1 | PDB_04-01-2025 0 | KEGG_04-01-2025 0 |
| 01-01-2025 | SwissProt_01-01-2025 3 | PDB_01-01-2025 2 | KEGG_01-01-2025 1 |
| 11-23-2024 | SwissProt_11-23-2024 0 | PDB_11-23-2024 0 | KEGG UPDATE HISTORY |
| 09-26-2024 | SwissProt_09-26-2024 0 | PDB_09-26-2024 0 | KEGG UPDATE HISTORY |
| 07-22-2024 | SwissProt_07-22-2024 2 | PDB_07-22-2024 1 | KEGG UPDATE HISTORY |
| 05-15-2024 | SwissProt_05-15-2024 0 | PDB_05-15-2024 0 | KEGG UPDATE HISTORY |
| 03-13-2024 | SwissProt_03-13-2024 0 | PDB_03-13-2024 0 | KEGG UPDATE HISTORY |
| 01-08-2024 | SwissProt_01-08-2024 1 | PDB_01-08-2024 1 | KEGG UPDATE HISTORY |
Overview model performances
- Training and model configuration
The ECPICK, HIT-EC, and CLEAN models were all trained using the same curated dataset. For the three models, we adopted the hyperparameters exactly as specified in their original publication. It is important to note that no separate validation or test sets were used during the training phase. - Performance evaluation
To evaluate the models, we calculated the micro-averaged F1 score and macro-averaged F1 score. Although the evaluation data was included in the training process, the performance metrics were strictly based on samples from the overlapping classes between the original databases (Swiss-Prot, PDB, KEGG) and the curated dataset.