Table 4:

Performance evaluation results of cross-validation tests on Gram-positive bacteria dataset for different predictions: pseudo-amino acid composition (PseAAC), position-specific scoring matrix (PSSM) profiles, gene ontology (GO) terms 0/1-based representation, GO terms PPV-based representation, features fusion, a consensus prediction using both GO terms and PSSM profiles outputs, and a consensus of PseAAC, GO terms and PSSM profiles outputs. Italic values correspond to the best predictive model.

Sequence features	Example-based metrics							Label-based metrics
	Accuracy	Precision	Recall	F1 score	Subset_accuracy	Hamming-loss	Rank-loss	Macro_precision	Macro_recall	Macro_F1 score	Micro_precision	Micro_recall	Micro_F1 score
PseACC	0.896	0.897	0.895	0.896	0.894	0.041	0.104	0.708	0.510	0.559	0.897	0.894	0.895
PSSM profiles	0.937	0.938	0.937	0.937	0.935	0.025	0.062	0.909	0.763	0.812	0.938	0.935	0.937
GO terms 0/1	0.959	0.959	0.958	0.959	0.957	0.016	0.041	0.952	0.785	0.791	0.959	0.957	0.958
GO terms ppv	0.962	0.962	0.961	0.962	0.960	0.015	0.038	0.927	0.803	0.819	0.962	0.960	0.961
PseAAC+PSSM+GO	0.947	0.949	0.947	0.948	0.946	0.020	0.052	0.931	0.772	0.824	0.949	0.946	0.947
PSSM+GO	0.948	0.950	0.948	0.949	0.946	0.020	0.051	0.920	0.822	0.851	0.950	0.947	0.948
ConsensusPseAAC+PSSM+GO	0.924	0.924	0.978	0.941	0.871	0.033	0.050	0.855	0.858	0.839	0.871	0.977	0.921
ConsensusPSSM+GO	0.950	0.950	0.974	0.958	0.924	0.021	0.038	0.886	0.854	0.851	0.924	0.973	0.948