Figure 5

Epigenetic signature of active and bivalent genes from mESCs reconstructed with SeqCode. (a) Bivalent region repressed for transcription by PcG proteins containing the HoxD complex. Genome-wide profiles and peaks of H3K4me3 and H3K27me3 are shown along this locus. (b) Overlap of H3K4me3 and H3K27me3 peaks produces three classes of sites: active (H3K4me3 + /H3K27me3–; orange), bivalent (H3K4me3 + /H3K27me3 + ; violet), and silent (H3K4me3–/H3K27me3 + ; blue). Determination of signal strength and genomic distribution of each class of peaks. (c) Overlap of H3K4me3 and H3K27me3 target genes, showing the functional analysis of active genes (orange), bivalent genes (violet), and silent genes (blue). (d) Average distribution and heatmap of the ChIP-seq signal around the TSS of active and bivalent genes for H3K4me3, H3K27me3, H3K36me3, and Suz12. (e) ChIP-seq levels of each experiment are shown for active and bivalent genes (left and right boxes in the boxplots, respectively). For comparison, gene expression (in RPKMs) is shown for both gene sets. Raw data of ChIP-seq and RNA-seq experiments were retrieved from³⁶. The SeqCode buildChIPprofile function was used to generate each custom track from the resulting BAM files after mapping; the matchpeaks application compared the H3K4me3 and H3K27me3 peaks; the recoverChIPlevels application determined the strength of the ChIP-seq signal at each subset of peaks; the genomeDistribution program calculated the genomic composition of each collection of peaks, according to RefSeq annotations; the matchpeaksgenes routine associated ChIP-seq peaks and target genes; and the produceTSSplots and produceTSSmaps programs generated the average distribution meta-plots and heatmaps of each ChIP-seq sample for signal around the TSS of active and bivalent genes, respectively. GO term enrichment was analyzed with Enrichr⁸⁰. For boxplots in (b,e), ChIP-seq counts were normalized by the total number of mapped reads, and RNA-seq expression values were calculated as RPKMs.