We further tested whether the expression level for each sample and candidate enhancer was significantly greater than the genomic background (see construction of random genomic background regions below). A P-value was calculated for each enhancer expression value for each primary cell, tissue and cell line sample by counting the fraction of random genomic regions with greater expression level in the same sample. Enhancers with P-values less than 0.001 and Benjamini-Hochberg adjusted FDR < 0.05 was considered transcribed in that sample. This analysis yielded binary expression values, which were used for constructing enhancer sets associated with each sample. In total, 38,554 enhancers were transcribed at a significant expression level in at least one primary cell or tissue sample. Below, we refer to this set as the ‘robust set’ of enhancers and indicate whenever it was used. For all analyses, we use the whole (‘permissive’) set of 43,011 enhancers if not otherwise mentioned.
