Table 8 Ablation results for different steps of the AMPDeep pipeline on the XGBC-Hem dataset
From: AMPDeep: hemolytic activity prediction of antimicrobial peptides using transfer learning
Approach | MCC |
|---|---|
Baseline | 0.846 |
BERT Init. (Natural Language) + Classification Layer Fine-Tuning | 0.6639 |
BERT Init. (Natural Language) + Mean Pooling + Classification Layer Fine-Tuning | 0.8091 |
Prot-BERT Init. + Mean Pooling + Classification Layer Fine-Tuning | 0.8472 |
Prot-BERT Init. + Mean Pooling + Selective Fine-Tuning | 0.9367 |
Prot-BERT + Mean Pool. + Secretion Transfer Learning + Selective Fine-Tun. (AMPDeep) | 0.9731 |