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

Table 11 Comparison of our predictors with other models via 10-fold cross-validation for D. melanogaster

Method	Feature	ACC	Sn	Sp	MCC	AUC
DLNN-5 [24]		0.8560	0.8781	0.8333	~	~
ZCMM [23]		0.9362	0.9226	0.7964	0.7000	0.9110
SVM	FCGR	0.8113	0.7831	0.84	0.6241	0.8791
	FCGR + DAC	0.8089	0.7835	0.8347	0.619	0.8747
	FCGR + TAC	0.8115	0.7831	0.8404	0.6245	0.8766
	FCGR + DACC	0.7809	0.7931	0.7684	0.5621	0.8343
	FCGR + TACC	0.7678	0.6879	0.8491	0.544	0.8363
	FCGR + PCPseDNC	0.8085	0.7752	0.8425	0.619	0.8773
	FCGR + PCPseTNC	0.8097	0.7772	0.8428	0.6215	0.8761
	All features	0.7593	0.70414	0.81544	0.52275	0.80283
ELM	FCGR	0.791	0.7648	0.8175	0.5833	0.8595
	FCGR + DAC	0.792	0.7779	0.8063	0.5847	0.8644
	FCGR + TAC	0.7917	0.7807	0.8028	0.5839	0.8651
	FCGR + DACC	0.7769	0.7617	0.7923	0.5544	0.8503
	FCGR + TACC	0.7694	0.7735	0.7653	0.5391	0.846
	FCGR + PCPseDNC	0.7896	0.7631	0.8165	0.5806	0.8651
	FCGR + PseTNC	0.7678	0.7283	0.8081	0.5379	0.8448
	All features	0.7847	0.781	0.7884	0.57	0.8576
XGBoost	FCGR	0.8037	0.7824	0.8253	0.6085	0.8772
	FCGR + DAC	0.7877	0.7683	0.8074	0.5763	0.863
	FCGR + TAC	0.793	0.7635	0.8232	0.5879	0.8671
	FCGR + DACC	0.7741	0.769	0.7793	0.5486	0.8506
	FCGR + TACC	0.7824	0.7693	0.7958	0.5658	0.8542
	FCGR + PCPseDNC	0.7997	0.7824	0.8172	0.6001	0.8725
	FCGR + PseTNC	0.7988	0.779	0.819	0.5989	0.8775
	All features	0.7951	0.7793	0.8112	0.5909	0.8718
MLP	FCGR	0.8117	0.8000	0.8235	0.6238	0.8848
CNN	FCGR	0.8108	0.8014	0.8204	0.6228	0.8854

The table shows the optimal results of each classifier, and the specific parameters are shown in the previous tables
Best values are in bold

ISSN: 1471-2105