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

Table 3 The prediction results of integrative feature representation for C. elegans 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	0.8574	0.8863	0.8290	0.7164	0.9283
	FCGR + TAC	K = 4, lag = 2	0.8561	0.8824	0.8302	0.7137	0.9272
	FCGR + DACC	K = 4, lag = 2	0.8471	0.8777	0.8171	0.6961	0.9122
	FCGR + TACC	K = 4, lag = 2	0.8470	0.8641	0.8301	0.6949	0.9179
	FCGR + PCPseDNC	K = 4, λ = 8, w = 0.5	0.8576	0.8921	0.8236	0.7176	0.9275
	FCGR + PCPseTNC	K = 4, λ = 8, w = 0.5	0.8539	0.8839	0.8244	0.7096	0.9275
	All features		0.8431	0.8461	0.8401	0.6867	0.9139
ELM	FCGR + DAC	K = 4, lag = 2	0.8707	0.8863	0.8555	0.7421	0.9355
	FCGR + TAC	K = 4, lag = 2	0.8696	0.8890	0.8505	0.7400	0.9359
	FCGR + DACC	K = 4, lag = 2	0.8684	0.8831	0.8539	0.7376	0.9358
	FCGR + TACC	K = 4, lag = 2	0.8680	0.8917	0.8447	0.7371	0.9329
	FCGR + PCPseDNC	K = 4, λ = 8, w = 0.5	0.8624	0.8847	0.8405	0.7258	0.9318
	FCGR + PCPseTNC	K = 4, λ = 8, w = 0.5	0.8557	0.8847	0.8271	0.7132	0.9262
	All features		0.8597	0.8863	0.8336	0.7210	0.9271
XGBoost	FCGR	K = 1 + 2 + 4	0.8487	0.8797	0.8182	0.6995	0.9202
	FCGR + DAC	K = 4, lag = 2	0.8416	0.8652	0.8182	0.6842	0.9165
	FCGR + TAC	K = 4, lag = 2	0.8433	0.8707	0.8163	0.6882	0.9169
	FCGR + DACC	K = 4, lag = 2	0.8462	0.8703	0.8225	0.6938	0.9170
	FCGR + TACC	K = 4, lag = 2	0.8417	0.8676	0.8163	0.6848	0.9162
	FCGR + PCPseDNC	K = 4, λ = 8, w = 0.5	0.8450	0.8749	0.8156	0.6917	0.9199
	FCGR + PCPseTNC	K = 4, λ = 8, w = 0.5	0.8493	0.8789	0.8202	0.7004	0.9178
	All features		0.8481	0.8695	0.8271	0.6973	0.9195

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