DPI_CDF: druggable protein identifier using cascade deep forest

BMC Bioinformatics

Table 5 Performance of various feature descriptors on independent dataset DPind

Classifier	Feature vector	ACC (%)	SEN (%)	SPE (%)	MCC	AUC
MLP	CPSR	89.37	83.03	95.35	0.792	0.940
	NQLC	90.23	86.60	93.67	0.805	0.937
	HOG-PSSM	78.09	73.66	82.27	0.562	0.845
	Hybrid1	86.98	91.51	82.70	0.743	0.940
	Hybrid2	88.93	86.16	91.56	0.779	0.936
	Hybrid3	89.37	88.39	90.29	0.787	0.946
ERT	CPSR	86.55	79.46	93.24	0.736	0.892
	NQLC	83.08	75.00	90.71	0.667	0.880
	HOG-PSSM	77.87	73.66	81.85	0.557	0.838
	Hybrid1	72.45	70.08	74.68	0.448	0.799
	Hybrid2	77.86	74.55	81.01	0.557	0.865
	Hybrid3	75.92	68.75	82.70	0.520	0.855
XGBoost	CPSR	87.85	83.03	92.40	0.759	0.912
	NQLC	87.85	82.14	93.24	0.760	0.917
	HOG-PSSM	89.37	88.83	89.87	0.787	0.964
	Hybrid1	89.15	88.83	89.45	0.782	0.936
	Hybrid2	89.37	88.83	89.87	0.787	0.941
	Hybrid3	89.15	88.83	89.45	0.782	0.936
DPI_CDF	CPSR	87.41	87.94	86.91	0.748	0.927
	NQLC	85.24	79.91	90.29	0.707	0.893
	HOG-PSSM	94.14	96.42	91.98	0.883	0.980
	Hybrid1	95.01	96.87	93.24	0.900	0.986
	Hybrid2	94.36	96.86	91.98	0.888	0.977
	Hybrid3	94.57	96.87	92.40	0.892	0.978

ISSN: 1471-2105