Table 1.

Performance metrics of three different deep representation learning features using three machine learning models.

Feature	Model	Dim	10-Fold Cross-Validation							Independent Test
			ACC	MCC	Sn	Sp	F1	auPRC	auROC	ACC	MCC	Sn	Sp	F1	auPRC	auROC
SSA b	SVM c	121	0.826	0.652	0.836	0.816	0.828	0.89	0.898	0.883 a	0.766	0.891	0.875	0.884	0.951	0.944
	LGBM c		0.787	0.575	0.816	0.758	0.793	0.874	0.886	0.859	0.722	0.906	0.812	0.866	0.949	0.941
	RF c		0.791	0.584	0.828	0.754	0.798	0.848	0.865	0.82	0.644	0.875	0.766	0.83	0.934	0.922
UniRep b	SVM c	1900	0.865	0.73	0.867	0.863	0.865	0.937	0.931	0.867	0.735	0.844	0.891	0.864	0.952	0.948
	LGBM c		0.84	0.68	0.828	0.852	0.838	0.939	0.93	0.867	0.735	0.844	0.891	0.864	0.953	0.952
	RF c		0.842	0.684	0.836	0.848	0.841	0.927	0.92	0.844	0.688	0.828	0.859	0.841	0.946	0.943
BiLSTM b	SVM c	3605	0.818	0.637	0.82	0.816	0.819	0.91	0.912	0.883	0.766	0.906	0.859	0.885	0.956	0.951
	LGBM c		0.855	0.711	0.863	0.848	0.857	0.924	0.926	0.836	0.673	0.812	0.859	0.832	0.95	0.95
	RF c		0.818	0.637	0.828	0.809	0.82	0.9	0.908	0.844	0.688	0.844	0.844	0.844	0.954	0.949

^a Best performance values are in bold and are underlined. ^b SSA: soft symmetric alignment; UniRep: unified representation; BiLSTM: bidirectional long short-term memory. ^c SVM: support vector machine; LGBM: light gradient boosting machine; RF: random forest.