In order to cluster regulatory (i.e., enhancer, promoter or dyadic) regions based on their activity patterns across all reference epigenomes, we expressed each region in terms of a binary vector of length nxs, where n is the number of reference epigenomes (111) and s is the number of chromatin states considered. For enhancers and promoters, s=3, as both of these types of regions are made up of 3 chromatin states in the 15-state ChromHMM model (enhancers: EnhG, Enh & EnhBiv, promoters: TssA, TssAFlnk & TssBiv).
