Extended Data Fig. 12: Hc-P enrichment at disease-relevant TF genes.

a, Enrichment of ‘molecular function’ GO terms in genes near repressed regions (state 13, Hc-P) as measured by GREAT binomial test with Benjamini–Hochberg correction. GO terms on the y-axis are ordered by average enrichment P value across all tissue-stages. The top 20 GO terms are listed below, and are all related to TF function. Number of regions for each tissue-stage shown in Extended Data Fig. 11e. b, Similar layout to Fig. 2f. The fractions of six gene sets that show evidence of PcG repression are plotted: 1) all protein-coding genes (black line); 2) the subset of protein-coding genes that code for TFs (green line); 3) the subset of protein-coding genes that code for TFs and underlie human Mendelian diseases (dark blue line); 4) the subset of protein-coding genes that code for TFs but do not underlie human Mendelian diseases (light blue line); 5) the subset of protein-coding genes that underlie human Mendelian diseases; 6) the subset of protein-coding genes that underlie human Mendelian diseases but are not TFs. The origin of the TF super-sets is indicated on top of each sub-panel, from left to right: the TFClass database, the DBD database, and genes associated with a GO term containing the phrase ‘TF’. c, P values from χ² test of independence between PcG repression and Mendelian phenotype involvement. Different subsets of TF genes were used for this analysis, clockwise from top to bottom: All, all genes annotated as TF in the indicated database (TFClass or DBD); non-Zf, genes annotated as TF but not as zinc finger, to ensure that the enrichment for disease genes is not coming only from this large family of TFs; GO term development, genes with a GO term containing ‘development’, to show that the enrichment for disease genes exists even amongst TFs that are all likely to have a role in development; CCDS, genes with transcripts annotated by the consensus coding sequence (CCDS) project, representing high-confidence gene annotations in both the mouse and human genomes. Sample sizes shown over each bar. d, Patterns of PcG repression at Sox9 (chr11: 112,766,260–112,803,708; mm10), Shh (chr5: 28,392,703–28,531,239; mm10), Pax3 (chr1: 78,027,730–78,280,060; mm10), and Wnt6/Ihh (chr1: 74,643,751–74,987,517; mm10). This small but well-characterized set of genes is known to cause human congenital phenotypes when expressed ectopically during development^46,96.