peaks, and we required adjacent peaks to be at least 500 bp away to avoid redundant detection. Only one tag from each unique position was considered to avoid clonal artifacts from sequencing. The threshold for the number of tags that determined a valid peak was selected at a false discovery rate of 0.001 determined by peak finding using randomized tag positions in a genome with an effective size of 2 × 109 bp. We also required peaks to have at least 4-fold more tags (normalized to total count) than input control samples. In addition, we required 4-fold more tags relative to the local background region (10 kb) to avoid identifying regions with genomic duplications or nonlocalized binding. Annotated positions for promoters, exons, introns and other features were based on RefSeq transcripts and repeat annotations from the University of California, Santa Cruz. Peaks from separate experiments were considered equivalent/co-bound if their peak centers were located within 200 bp of each other. Read density heat maps were created by first using HOMER to generate read densities and then visualized using Java TreeView (http://jtreeview.sourceforge.net). To define the overlapping of two ChIP-seq, for instance of OLIG2 and H3K9me3, we first found out OLIG2 binding
