Comparative analysis and prediction of nucleosome positioning using integrative feature representation and machine learning algorithms

Table 12 Comparison of our predictors with other advanced models via 20-fold cross-validation for H. sapiens

Method	Feature	ACC	Sn	Sp	MCC	AUC
LeNup [25]		0.8889	0.9212	0.8562	0.7906	0.9412
3LS [22]		0.9001	0.9169	0.8835	0.8006	0.9588
SVM	FCGR	0.8760	0.8940	0.8583	0.7535	0.9288
	FCGR + DAC	0.8751	0.8874	0.8630	0.7513	0.9310
	FCGR + TAC	0.8754	0.8869	0.8639	0.7519	0.9318
	FCGR + DACC	0.8563	0.8544	0.8583	0.7138	0.9217
	FCGR + TACC	0.8423	0.8337	0.8509	0.6858	0.9114
	FCGR + PseDNC	0.8736	0.8883	0.8591	0.7481	0.9294
	FCGR + PseTNC	0.8740	0.8905	0.8578	0.7491	0.9280
	All features	0.8154	0.7545	0.8757	0.6355	0.8998
ELM	FCGR	0.8456	0.8702	0.8213	0.6931	0.9092
	FCGR + DAC	0.8469	0.8707	0.8235	0.6952	0.9087
	FCGR + TAC	0.8478	0.8750	0.8209	0.6974	0.9142
	FCGR + DACC	0.8500	0.8772	0.8230	0.7017	0.9054
	FCGR + TACC	0.8454	0.8865	0.8048	0.6941	0.9104
	FCGR + PseDNC	0.8476	0.8640	0.8313	0.6968	0.9111
	FCGR + PseTNC	0.8439	0.8702	0.8178	0.6893	0.9111
	All features	0.8474	0.8909	0.8044	0.6980	0.9141
XGBoost	FCGR	0.8602	0.897	0.8239	0.7235	0.9237
	FCGR + DAC	0.8561	0.8627	0.8496	0.7130	0.9186
	FCGR + TAC	0.8865	0.9035	0.8696	0.7734	0.9394
	FCGR + DACC	0.8439	0.8667	0.8213	0.6894	0.9136
	FCGR + TACC	0.8406	0.8711	0.8104	0.6831	0.9046
	FCGR + PseDNC	0.8563	0.8931	0.82	0.7152	0.9208
	FCGR + PseTNC	0.8504	0.8712	0.8300	0.7029	0.9185
	All features	0.8511	0.8755	0.827	0.7039	0.9193
MLP	FCGR	0.8579	0.8839	0.8322	0.7172	0.9186
CNN	FCGR	0.8616	0.8746	0.8487	0.7239	0.9222

ISSN: 1471-2105