Figure 1

Genome-wide histograms of ChIP-chip probe activity can be decomposed into different clusters with distinct shapes. (A) Illustrative example showing acetylations, promoters and genes in ENCODE region ENr333. Riken H3K9ac: signal strength of H3K9ac, determined by ChIP-chip at Riken. SI H3ac GM06990: H3 lysine 9/14 acetylation (H3ac) signal from ENCODE [15] for comparison. Riken DeepCAGE: promoters in the same THP-1 cell line as Riken H3K9ac. RefSeq Genes: RefSeq Genes in this region. Notably, the H3K9ac and H3ac signal are localized at promoters and RefSeq start sites. (B) Genome-wide histogram of the H3K9ac level of 4,481 Fantom 4 promoters. The promoters have been aligned at their representative position (see methods). The x-axis show the genomic region +/-2,200 bp around all promoters. The y-axis shows the average H3K9 acetylation signal (the -10log(p-value) of H3K9ac) in bins of size 400 bp (see methods section for details). (C)(D)(E) Examples of individual core promoters in the upstream, centered and downstream clusters, showing the surrounding H3K9ac signal in the +/-2,200 region around the representative position (red line) without binning. The red line indicates the representative position of the promoter. (F) The 4,481 promoters have been clustered according to the Kolmogorov-distance of the normalized, cumulative ChIP-chip probe activity in the +/- 2,000 bp region around each promoter, using k-medoids clustering. The number of promoters in each cluster is shown in the legend. The three histograms visualizing each cluster are drawn with the same parameters as the histogram in (B).