Extended Data Fig. 11: Properties of putative PcG target genes.

a, TSSs are binned together according to the number of tissue-stages in which they are marked by Hc-P (0–66, x-axis). For each bin, the fraction of TSSs that are K4 + K27 (bivalent), K27 (repressed), K4 (active), or has no K4 or K27 in mouse ES cells is plotted, as reported previously²⁹. b–d, Similar schema to a, but plotting the fraction of TSSs bound by RING1B (PRC1 component), EZH2 (PRC2 component), or SUZ12 (PRC2 component) in mouse ES cells, as previously reported³⁰. e, Comparison of Hc-P regions as reported here and DMVs from ref. ⁷. Left, metrics related to regions annotated as Hc-P in each tissue-stage (x-axis). From top to bottom: number of Hc-P regions in each tissue-stage; coverage of Hc-P in each tissue-stage; fraction of Hc-P regions that overlap a TSS; fraction of Hc-P regions that overlap a DMV. Right, metrics related to regions annotated as DMVs in each tissue-stage (x-axis). From top to bottom: number of DMVs in each tissue-stage; coverage of DMVs in each tissue-stage; fraction of DMV regions that overlap a TSS; fraction of DMV regions that overlap a Hc-P region. f, Schema as in a–d, but with axes switched. For each bin, the fraction of TSSs that overlap a CGI is plotted on the x-axis. g–j, The following properties of CGIs that overlapped Hc-P TSSs are plotted (left to right): CGI length; CpG number; CPG percentage; GC percentage. None of these properties is strongly correlated with the number of tissue-stages in which a given TSS is marked by Hc-P (x-axis), supporting the role of factors other than CGIs in recruiting or excluding PcG at target promoters in a tissue- and/or stage-restricted fashion^94,95. Green line shows LOESS smooth curve, span 0.25 and degree 1.