Table 4 Evaluation of ML classifiers’ performance utilizing various PLM embedding sources on an independent dataset

From: Prediction of hemolytic peptides and their hemolytic concentration

Embedding source

ML classifier

Sp (%)

Sn (%)

Acc (%)

MCC

AUC

ESM2-t36

XGBC

86.0

74.9

80.8

0.614

0.869

ESM2-t33

ET

85.5

73.2

79.8

0.604

0.873

ESM2-t6

ET

73.2

84.5

79.3

0.583

0.857

ProtBERT

MLPC

83.9

84.0

83.9

0.677

0.882

BioBERT

MLPC

81.5

68.5

75.6

0.507

0.792

ProtBERT+BiLSTM

ET

82.8

71.4

77.4

0.547

0.859

  1. Sn sensitivity, Sp specificity, Acc accuracy, MCC Matthews correlation coefficient, AUC area under receiver operating characteristic, Bold values indicate the best-performing model.
Back to article page