ChIP-seq analysis

Sequencing reads were aligned against the hg19 reference genome using BWA (Li and Durbin, 2010). Alignments were filtered for uniquely mapped reads and duplicates were removed. To determine the regions of ChIP-Seq tag enrichment for broad histone marks (H3K9me1, H3K9me2, H3K9me3, H3K36me1, H3K36me2, H3K36me3, H3K27me3), we analyzed tag counts in 50 Kb bins across the chromosome length. All counts in both replicates were quantile normalized across both control and KDM4A overexpression datasets in all phases (G1/ES/LS/G2). These normalized counts were then used to calculate ChIP enrichment over input as the log₂ ratio of ChIP to input tag density. Peaks of enrichment for focused histone marks (H3K4me1, H3K4me2, H3K4me3, H3K27ac, H3K9ac) were called using HOMER (Heinz et al., 2010). To identify differential ChIP-seq enrichment between control and KDM4A overexpressing cells, we analyzed input-normalized ChIP-seq tag densities at either all 50 Kb genomic bins for broad marks, or all regions corresponding to the union of peaks called in individual samples for focused marks, and used edgeR (Robinson et al., 2010) with the cutoff of at least 1.5-fold difference between replicate averages. Public ChIP-seq data were downloaded from GEO ({"type":"entrez-geo","attrs":{"text":"GSE73696","term_id":"73696"}}GSE73696, {"type":"entrez-geo","attrs":{"text":"GSE118954","term_id":"118954"}}GSE118954). To compare our RPE data on H3K36me2 enrichment to other cell types, H3K36me2-enriched regions were defined as 50 Kb bins with at least 1.5-fold ChIP enrichment over input, followed by determining the overlaps between these regions in different datasets and estimating their statistical significance as Z-scores and P values based on the distribution of overlap lengths between randomly shuffled regions.