Pooled CAGE data from all FANTOM5 libraries (described above) were further overlaid with these regions and a directionality score based on the aggregate of CAGE tags falling within +/− 300bp from the center positions were calculated to determine potential strand bias. For comparison, we repeated the same calculations for genomic regions +/− 300bp around TSSs of RefSeq protein coding genes. Directionality was calculated as (F − R) / (F + R), where F and R is the sum of CAGE tags aligned on the forward and reverse strand, respectively. Directionality close to −1 or 1 indicates a unidirectional behavior while 0 indicates perfectly balanced bidirectional transcription.
