Fig. 3

Prediction of CRMs using different \({S}_{CRM}\) cutoffs. A A cartoon shows the proportions of the 79.9% of genome regions covered by originally called binding peaks (64.2%) and their extended parts (35.8%) as well as their relative contributions to the predicted CRMs (kept original (69.5%) and kept extended (30.5%)) and non-CRMCs (abandoned original (51.4%) and abandoned extended (41.3%)). Percentage above the lines are the proportion of originally called binding peaks and their extended parts that are predicted to be CRMCs and non-CRMCs. B Distribution of the \({S}_{CRM}\) scores of the CRMCs and the Null CRMCs. The inset is a blowup view of the indicated regions. The dotted vertical lines indicate \({S}_{CRM}\) cutoffs for the corresponding p-values. C Number of the predicted CRMs, proportion of the genome predicted to be CRMs and the corresponding p-value as functions of the \({S}_{CRM}\) cutoff α. D Percentage of the genome that are predicted to be CRM and TFBS positions in exonic sequences (ESs) and non-exonic sequences (NESs) using various \({S}_{CRM}\) cutoffs and corresponding p-values. E Percentage of NESs that are predicted to be CRMs and TFBSs using various \({S}_{CRM}\) cutoffs and corresponding p-values. F Distribution of the lengths of CRMs predicted using different \({S}_{CRM}\)  cutoffs and corresponding p-values