Co-AMPpred for in silico-aided predictions of antimicrobial peptides by integrating composition-based features

BMC Bioinformatics

Table 1 Performances of machine-learning models on the benchmark training and independent test datasets. Values shown are mean ± SD for the training dataset

Algorithm	Dataset	Accuracy	AUROC	Recall	Precision	Kappa	MCC
GBC	Training	75.0% ± 0.038	0.816 ± 0.035	77.4% ± 0.082	73.9% ± 0.033	0.500 ± 0.0755	0.504 ± 0.075
GBC	Test	80.3%	0.873	79.7%	80.6%	0.606	0.606
CatBoost	Training	74.4% ± 0.055	0.815 ± 0.045	75.3% ± 0.107	73.9% ± 0.045	0.488 ± 0.110	0.492 ± 0.109
CatBoost	Test	78.7%	0.879	78.7%	78.7%	0.574	0.574
LGBM	Training	73.8% ± 0.060	0.810 ± 0.052	73.3% ± 0.124	73.8% ± 0.039	0.476 ± 0.102	0.479 ± 0.099
LGBM	Test	77.6%	0.868	78.7%	77.0%	0.553	0.553
ETC	Training	74.3% ± 0.055	0.794 ± 0.066	75.0% ± 0.097	73.9% ± 0.049	0.487 ± 0.109	0.491 ± 0.108
ETC	Test	77.6%	0.776	77.6%	77.6%	0.553	0.553
RF	Training	74.1% ± 0.044	0.798 ± 0.052	75.5% ± 0.101	73.1% ± 0.039	0.482 ± 0.088	0.487 ± 0.086
RF	Test	78.1%	0.811	78.7%	77.8%	0.563	0.563

The given data in bold font indicates the top performance of the model on the test dataset
GBC, gradient boosting classifier; LGBM, light gradient boosting machine; ETC, extra trees classifier; RF, random forest; AUROC, area under the receiver operating characteristics curve; MCC, Mathew's correlation coefficient; SD, standard deviation

ISSN: 1471-2105