Identification of super-enhancers

Super-enhancers were identified using the rank ordering of super-enhancers (ROSE) algorithm (http://younglab.wi.mit.edu/super_enhancer_code.html, [31]). Briefly, peaks were called on H3K27ace ChIP-Seq bam files using MACS2 as described above using the human reference genome GRCh37/hg19 (see peak-calling procedure). Peaks located within 2500 base pairs (bp) of TSSs were excluded, as a 5-kb TSS exclusion zone was applied. The remaining peaks were designated as putative enhancers. Enhancers located within 12,500 base pairs of each other were stitched together, scored, and ranked based on the H3K27ac ChIP-Seq signal (total normalized read count in the ChIP-Seq sample after input subtraction). Then, enhancer signal (H3K27ac density) was plotted against enhancer rank. Enhancer regions above the inflection point of the curve were designated as super-enhancers, and those below the inflection point were designated as typical enhancers. Super-enhancers were assigned to the nearest RefSeq gene or genes using the 3 default parameters of the ROSE algorithm: included were overlapping genes (transcript directly overlapping the super-enhancer), proximal genes (TSS within 50 kb of the center of the super-enhancer), and nearest gene.