DPI_CDF: druggable protein identifier using cascade deep forest

BMC Bioinformatics

Table 4 Performance of various feature descriptors on DPtrain benchmark dataset using 10-fold CV test

Classifier	Feature vector	ACC (%)	SEN (%)	SPE (%)	MCC	AUC
MLP	CPSR	88.08	86.63	86.63	0.770	0.932
	NQLC	87.14	87.27	87.04	0.750	0.958
	HOG-PSSM	79.23	85.69	73.24	0.594	0.872
	Hybrid1	88.20	87.85	88.56	0.776	0.950
	Hybrid2	87.69	85.31	89.92	0.760	0.954
	Hybrid3	89.98	86.63	93.10	0.810	0.953
ERT	CPSR	87.34	85.47	89.09	0.751	0.934
	NQLC	86.08	83.91	88.1	0.724	0.928
	HOG-PSSM	82.06	83.89	80.37	0.643	0.878
	Hybrid1	80.17	82.00	78.48	0.605	0.883
	Hybrid2	83.87	83.41	84.31	0.678	0.920
	Hybrid3	83.16	82.99	83.33	0.664	0.916
XGBoost	CPSR	87.77	85.63	89.76	0.759	0.944
	NQLC	88.71	85.88	91.35	0.778	0.949
	HOG-PSSM	93.63	94.10	93.18	0.876	0.986
	Hybrid1	93.62	93.20	94.01	0.873	0.968
	Hybrid2	92.76	93.20	92.34	0.859	0.969
	Hybrid3	93.59	93.21	93.93	0.873	0.968
DPI_CDF	CPSR	90.09	87.52	92.49	0.806	0.956
	NQLC	89.22	85.96	92.26	0.788	0.949
	HOG-PSSM	94.77	93.64	95.82	0.895	0.969
	Hybrid1	99.13	98.52	99.69	0.982	0.999
	Hybrid2	99.21	98.93	99.46	0.984	0.998
	Hybrid3	99.33	99.02	99.62	0.986	0.998

ISSN: 1471-2105