Genome-wide predictors of NF-κB recruitment and transcriptional activity

Table 1 Performance of epigenetic marks in the prediction of p65 binding

	Prox	Dist S1	Dist S2	mean	∆mot
h2az	0.72	0.61	0.60	0.64
h2az + mot	0.80	0.83	0.69	0.77	0.13
h3k27ac	0.70	0.64	0.61	0.65
h3k27ac + mot	0.80	0.84	0.69	0.78	0.13
h3k27me3	0.52	0.53	0.54	0.53
h3k27me3 + mot	0.70	0.80	0.66	0.72	0.19
h3k4me1	0.59	0.70	0.59	0.63
h3k4me1 + mot	0.74	0.86	0.68	0.76	0.14
h3k4me2	0.72	0.70	0.61	0.67
h3k4me2 + mot	0.80	0.86	0.70	0.79	0.11
h3k4me3	0.70	0.64	0.58	0.64
h3k4me3 + mot	0.80	0.84	0.68	0.77	0.13
h3k79me2	0.50	0.53	0.53	0.52
h3k79me2 + mot	0.69	0.80	0.65	0.71	0.20
h3k9ac	0.69	0.64	0.63	0.65
h3k9ac + mot	0.79	0.84	0.71	0.78	0.13
mot	0.69	0.79	0.63	0.71

Predictive performance (AUC – area under curve) of single-mark logistic regression models, with optional inclusion of binary NF-κB motif presence (+mot), on the binding of p65 at three classes of sites: Prox – TSS-proximal; Dist S1 – TSS-distal (stage 1); Dist S2 – TSS-distal (stage 2); mean – arithmetic average performance for all classes; ∆mot – arithmetic average improvement of the (+mot) models

ISSN: 1756-0381