. 2022 Jul 17;23(14):7877. doi: 10.3390/ijms23147877

Table 3.

Performance metrics of individual features and fused features, according to the machine learning methods used to derive them.

Feature	Model	Dim	10-Fold Cross-Validation							Independent Test
Feature	Model	Dim	ACC	MCC	Sn	Sp	F1	auPRC	auROC	ACC	MCC	Sn	Sp	F1	auPRC	auROC
SSA ^b	SVM ^c	53	0.820	0.641	0.840	0.801	0.824	0.910	0.909	0.914	0.829	0.937	0.891	0.916	0.948	0.941
	LGBM ^c	77	0.816	0.634	0.848	0.785	0.822	0.877	0.892	0.883	0.768	0.922	0.844	0.887	0.947	0.940
	RF ^c	16	0.805	0.610	0.820	0.789	0.808	0.860	0.881	0.867	0.734	0.875	0.859	0.868	0.888	0.894
UniRep ^b	SVM ^c	65	0.875	0.750	0.875	0.875	0.875	0.946	0.943	0.906	0.813	0.891	0.922	0.905	0.952	0.952
	LGBM ^c	313	0.854	0.707	0.855	0.852	0.854	0.946	0.938	0.914	0.829	0.891	0.937	0.912	0.954	0.948
	RF ^c	329	0.836	0.672	0.824	0.848	0.834	0.918	0.908	0.891	0.785	0.844	0.937	0.885	0.958	0.957
BiLSTM ^b	SVM ^c	344	0.820	0.641	0.824	0.816	0.821	0.913	0.915	0.922	0.844	0.937	0.906	0.923	0.955	0.956
	LGBM ^c	339	0.871	0.742	0.883	0.859	0.873	0.925	0.929	0.906	0.813	0.906	0.906	0.906	0.969	0.966
	RF ^c	434	0.830	0.660	0.836	0.824	0.831	0.906	0.914	0.898	0.797	0.906	0.891	0.899	0.957	0.950
SSA + UniRep ^b	SVM ^c	62	0.865	0.730	0.863	0.867	0.865	0.944	0.942	0.914	0.828	0.906	0.922	0.913	0.958	0.957
	LGBM ^c	106	0.881	0.762	0.887	0.875	0.882	0.961	0.957	0.891	0.783	0.859	0.922	0.887	0.952	0.947
	RF ^c	47	0.838	0.676	0.859	0.816	0.841	0.937	0.931	0.906	0.816	0.859	0.953	0.902	0.956	0.947
SSA + BiLSTM ^b	SVM ^c	267	0.836	0.672	0.836	0.836	0.836	0.910	0.911	0.914	0.828	0.906	0.922	0.913	0.956	0.952
	LGBM ^c	317	0.861	0.723	0.875	0.848	0.863	0.924	0.929	0.906	0.813	0.906	0.906	0.906	0.962	0.958
	RF ^c	176	0.832	0.664	0.848	0.816	0.835	0.922	0.925	0.906	0.813	0.906	0.906	0.906	0.959	0.952
UniRep + BiLSTM ^b	SVM ^c	186	0.873	0.746	0.887	0.859	0.875	0.932	0.934	0.914	0.829	0.937	0.891	0.916	0.961	0.965
	LGBM ^c	106	0.889 ^a	0.777	0.891	0.887	0.889	0.947	0.952	0.944	0.889	0.922	0.977	0.952	0.984	0.977
	RF ^c	45	0.871	0.742	0.871	0.871	0.871	0.937	0.941	0.938	0.875	0.938	0.938	0.938	0.976	0.971
SSA + UniRep + BiLSTM ^b	SVM ^c	336	0.881	0.762	0.883	0.879	0.881	0.940	0.942	0.922	0.845	0.953	0.891	0.924	0.942	0.946
	LGBM ^c	285	0.881	0.762	0.891	0.871	0.882	0.951	0.947	0.938	0.875	0.922	0.953	0.937	0.969	0.969
	RF ^c	192	0.863	0.727	0.859	0.867	0.863	0.932	0.932	0.922	0.844	0.906	0.937	0.921	0.970	0.967

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