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

Table 4 The prediction results of integrative feature representation for D. melanogaster via 10-fold cross-validation by SVM, ELM and XGBoost

Method	Feature	parameter	ACC	Sn	Sp	MCC	AUC
SVM	FCGR + DAC	K = 4, lag = 2	0.8047	0.7862	0.8235	0.6103	0.8762
	FCGR + TAC	K = 4, lag = 2	0.8089	0.7835	0.8347	0.6190	0.8747
	FCGR + DACC	K = 4, lag = 2	0.7753	0.7772	0.7733	0.5509	0.8295
	FCGR + TACC	K = 4, lag = 2	0.7560	0.6772	0.8361	0.5199	0.8247
	FCGR + PCPseDNC	K = 4, λ = 8, w = 0.5	0.8073	0.7797	0.8354	0.6162	0.8803
	FCGR + PCPseTNC	K = 4, λ = 8, w = 0.5	0.8057	0.7835	0.8284	0.6129	0.8769
	All features		0.7510	0.6828	0.8204	0.5078	0.7987
ELM	FCGR + DAC	K = 2, lag = 2	0.7920	0.7779	0.8063	0.5847	0.8644
	FCGR + TAC	K = 2, lag = 2	0.7917	0.7807	0.8028	0.5839	0.8651
	FCGR + DACC	K = 2, lag = 2	0.7769	0.7617	0.7923	0.5544	0.8503
	FCGR + TACC	K = 2, lag = 2	0.7694	0.7735	0.7653	0.5391	0.8460
	FCGR + PCPseDNC	K = 2, λ = 8, w = 0.5	0.7896	0.7631	0.8165	0.5806	0.8651
	FCGR + PCPseTNC	K = 2, λ = 8, w = 0.5	0.7595	0.7341	0.7853	0.5206	0.8400
	All features		0.7847	0.7810	0.7884	0.5700	0.8576
XGBoost	FCGR	K = 1 + 2 + 4	0.7976	0.7797	0.8158	0.5959	0.8725
	FCGR + DAC	K = 4, lag = 2	0.7873	0.7717	0.8032	0.5751	0.8613
	FCGR + TAC	K = 4, lag = 2	0.7877	0.7624	0.8133	0.5768	0.8647
	FCGR + DACC	K = 4, lag = 2	0.7724	0.7814	0.7632	0.5450	0.8532
	FCGR + TACC	K = 4, lag = 2	0.7824	0.7693	0.7958	0.5658	0.8542
	FCGR + PCPseDNC	K = 4, λ = 8, w = 0.5	0.7997	0.7824	0.8172	0.6001	0.8725
	FCGR + PCPseTNC	K = 4, λ = 8, w = 0.5	0.7988	0.7790	0.8190	0.5989	0.8775
	All features		0.7951	0.7793	0.8112	0.5909	0.8718

All features means the feature vector = FCGR + DACC + TACC + PC-PseDNC + PC-pseTAC, and the parameters are consistent with the parameters of the corresponding feature. Parameter K indicates the values of K nucleotide in FCGR; lag indicates the distance of lag along the sequence; λ represents the highest counted rank (or tier) of the correlation along a DNA sequence; w is the weight factor ranged from 0 to 1
Best values are in bold

ISSN: 1471-2105