Antimicrobial peptide prediction on BERT-based models benchmark leave-one-out 1.0
82ADAPTABLERDKit
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| RDKitclassifier=LightGBM, fingerprint=RDKit2025.01 | 82 | 93 | 88.4 | 91.4 | 76.4 | 92 | 87.2 | |
| AMP-BERTfine-tuned=false2025.01 | 81.7 | 78.3 | 82.3 | 62.7 | 64 | 82.8 | 75.3 | |
| ECFPclassifier=LightGBM, fingerprint=ECFP2025.01 | 80.8 | 93 | 88.2 | 91.5 | 76.3 | 95.8 | 87.6 | |
| TTclassifier=LightGBM, fingerprint=Topological Torsion2025.01 | 80.6 | 94.3 | 88.8 | 90.8 | 76.9 | 95.8 | 87.8 | |
| BERT-Proteinfine-tuned=false2025.01 | 79.4 | 90.1 | 80.6 | 82.9 | 58.8 | 58.8 | 75.1 | |
| LM_pred (BFD)fine-tuned=false2025.01 | 73.9 | 89.6 | 83.6 | 68.7 | 64.8 | 84.4 | 70.3 | |
| cAMPs predfine-tuned=false2025.01 | 61.3 | 74.1 | 72.8 | 52.4 | 39.2 | 85 | 64.1 | |
| LM predfine-tuned=false2025.01 | 55.9 | 71.7 | 64.5 | 60.7 | 47.6 | 82.1 | 63.8 |