Fig. 5

Exploration of master regulator TFs. a Leftmost heatmap indicates the cumulative probability of > X hits (probability of a hit is 1/20 in a 200-bp region) calculated as a Poisson distribution where the parameters are dependent on the number of active TFs in a cell and the numbers of TFs that specify a regulatory site. Remaining heatmaps show the effect of master regulators in the same framework in which probability in the leftmost heatmap is multiplied by the probability of getting a single master regulator (MR) hit [Poisson(1 MR hit | 1–4 MR available to the cell]. Cells with red shading are close to the expected probability of active regulatory elements in a cell type (i.e., ~ 0.5–3% of the genome). b AUROC curves for the trained LR models on DHS data in the ESC-H1 cell line using subset of the 1270 TF motifs identified using recursive feature selection (top 25, 50, 100, and 150 motifs). Models were also trained on motifs identified by gkm-Explain/TF-Modisco (9 motifs) and just the Yamanaka factors + CTCF (5 motifs). c Heatmaps indicating the number of 4-TF and 5-TF multimeric motif hits in the human genome, based on 3% cooperativity rate of dimers and the presence (5, 10, 20 MR) and absence of MR. Cells containing values between 30,000 and 200,000 are shaded red corresponding to the typical number of active regulatory elements in a cell type