A general integrative genomic feature transcription factor binding site prediction method applied to analysis of USF1 binding in cardiovascular disease

Table 2 Comparison of USF1-BS prediction models

Variables in the model	AUC	Standard error	Asymptotic 95% confidence interval		Sensitivity	Specificity
			Lower bound	Upper bound
PWM	0.648	0.053	0.544	0.752	0.176	0.989
PhastCons8	0.599	0.053	0.496	0.702	0.000	0.955
RP5	0.672	0.058	0.559	0.786	0.088	0.994
DNaseI	0.553	0.083	0.390	0.716	0.235	0.960
All features	0.639	0.060	0.522	0.756	0.382	0.898
Selected features	0.827	0.044	0.740	0.913	0.559	0.876

The threshold score that defined a predicted USF1-BS was 0.5. Sensitivity was the proportion of correctly predicted true positive loci, whereas specificity was the proportion of true negative loci predicted as negative loci. The AUC was calculated from LOLO cross-validation (see Methods).

ISSN: 1479-7364