Table 5.

The f1-scores of the compared methods and ARG-SHINE on the COALA90 test data for each class

ARG Class	BLAST best hit	DIAMOND best hit	DeepARG	HMMER	TRAC	ARG-SHINE
MULTIDRUG	0.8791	0.8696	0.8889	0.4268	0.7327	0.8842
AMINOGLYCOSIDE	0.8783	0.8929	0.8571	0.5382	0.8789	0.9099
MACROLIDE	0.9612	0.9612	0.9764	0.8552	0.9394	0.9767
BETA-LACTAM	0.8335	0.8299	0.8355	0.4518	0.8424	0.8584
GLYCOPEPTIDE	0.8284	0.8353	0.8451	0.6048	0.7637	0.8416
TRIMETHOPRIM	0.9630	0.9625	0.9434	0.6527	0.9068	0.9419
FSI	0.8578	0.8732	0.8683	0.1412	0.7443	0.8565
TETRACYCLINE	0.7923	0.7932	0.8241	0.6617	0.7824	0.8069
SULFONAMIDE	1.0000	1.0000	1.0000	0.8333	0.9836	1.0000
FOSFOMYCIN	0.7826	0.7826	0.9524	0.2778	0.7692	0.8696
PHENICOL	0.5366	0.5500	0.5823	0.2000	0.4675	0.5946
QUINOLONE	0.9804	0.7805	0.4375	0.8136	0.9600	0.9804
STREPTOGRAMIN	0.5000	0.5000	0.0000	0.0645	0.0000	0.5000
BACITRACIN	0.9524	1.0000	0.9524	0.5405	1.0000	1.0000
RIFAMYCIN	0.6667	0.6667	0.5000	0.0385	0.6667	0.6667
MLS	0.8000	0.6667	0.2222	0.0976	0.4000	1.0000
Macro-average	0.8258	0.8103	0.7303	0.4499	0.7399	0.8555
Weighted-average	0.8423	0.8423	0.8419	0.4916	0.8097	0.8591

‘Macro-average’ means that we average the f1-score of each individual class; ‘Weighted-average’ means that we average the f1-score of each individual class weighted by the number of sequences. The best values of f1-score for per class are in bold.