To compare motif signatures characterizing bidirectionally transcribed enhancers (permissive set) with those of CAGE-defined promoters, we used the set of 184,827 robust human CAGE clusters defined by 6 separated into 61,322 CGI and 123,505 nonCGI-associated clusters. We made further subsets of these CAGE clusters, contingent on their overlap with annotated TSSs from Refseq and Gencode. We merged overlapping extended CAGE clusters (−300, +50; based on the robust cluster set; average size nonCGI: 422 bp; average size CGI: 544 bp) contingent on CGI status and subtracted CAGE cluster regions that overlapped with extended enhancers (mid position +/− 200 bp).
