TSS and CpG island classification

The 5′-ends of GENCODE version 19 (ref. ⁴³) protein coding transcripts were extracted and ordered by the TSS coordinate, transcription direction, and the gene name. Transcripts with the same TSS, transcription direction and gene name were combined and the transcript with the highest number of RNA-seq reads in its first exon was chosen as a representative. This yielded 73,043 TSSs. We counted fragment centres in a window spanning ±750 base pairs from these TSSs for Input, H3K4me1, H3K4me3, H3K9me3, H3K27ac, H3K27me3, H3K36me3, H3K4me3 reChIP and H3K27me3 reChIP in R (ref. ⁴⁴) using the bioconductor package bamsignals (version 1.4 https://bioconductor.org/packages/bamsignals). For CpG islands, we downloaded the CpG island annotation from UCSC and counted reads along the complete length of the CpG island.

library(bamsignals)

# Genomic Ranges object tss or CpG island

gr = tss

count = bamCount(

bampath = bampath,

gr = gr,

mapqual = 20,

paired.end = “midpoint”,

tlen.filter = c(120,240),

filteredFlag = 1024

)

The (re)ChIP libraries were normalized against the appropriate control using normR:

library(normr)

norm = enrichR(

treatment = counts.(re)ChIP,

control = counts.control,

genome = gr

)

We called a TSS or CpG island enriched in a (re)ChIP using a q value threshold of 0.001, 0.01 and 0.1:

enriched.0.001 = !is.na(getClasses(norm, fdr = 0.001))

enriched.0.01 = !is.na(getClasses(norm, fdr = 0.01))

enriched.0.1 = !is.na(getClasses(norm, fdr = 0.1))

This procedure yields for every TSS or CpG island and for every (re)ChIP over control a classification, whether they are enriched or not. We used the enrichment calls for H3K4me3, H3K27me3, H3K4me3 reChIP and H3K27me3 reChIP over input to cluster the TSSs or CpG islands in eight classes: (i) no enrichment in any (re)ChIP, (ii) only enrichment in H3K27me3, (iii) only enrichment in H3K4me3, (iv) enrichment in H3K4me3 and H3K27me3 but not in the two reChIPs, (v) only enrichment in both the reChIPs but no enrichment in the primary ChIPs, (vi) enrichment in H3K27me3 and both the reChIPs, (vii) enrichment in H3K4me3 and both the reChIPs and (viii) enrichment in all the four (re)ChIPs.