Table 5 Performance of hybrid model on independent dataset developed by combining best-performing classification models with MERCI
From: Prediction of hemolytic peptides and their hemolytic concentration
Model | Threshold | Sp (%) | Sn (%) | Acc (%) | MCC | AUC | |
|---|---|---|---|---|---|---|---|
ML | RF (ALLCOMP-ex SOC) | 0.45 | 83.4 | 83.8 | 83.5 | 0.670 | 0.921 |
RF (AAC + DPC + PCP) | 0.49 | 82.6 | 77.6 | 83.9 | 0.621 | 0.899 | |
XGBC (ESM2t36 embeddings) | 0.47 | 84.0 | 80.8 | 82.9 | 0.658 | 0.900 | |
ET (ESM2t33 embeddings) | 0.49 | 85.6 | 82.4 | 82.8 | 0.651 | 0.905 | |
MLPC (ProtBERT embeddings) | 0.48 | 86.4 | 77.4 | 81.1 | 0.636 | 0.891 | |
ET (ESM2t6 embeddings) | 0.51 | 87.7 | 76.9 | 82.4 | 0.646 | 0.902 | |
PLM | ESM2-t33 | 0.57 | 92.7 | 68.5 | 83.5 | 0.668 | 0.901 |
ESM2-t30 | 0.48 | 79.7 | 82.7 | 80.2 | 0.634 | 0.885 | |
ESM2-t6 | 0.58 | 83.0 | 82.0 | 82.5 | 0.649 | 0.919 | |
ProtBERT | 0.44 | 65.9 | 90.3 | 79.0 | 0.585 | 0.871 | |
