Statistical significance of increased association between being a CRM and being transcribed over being a transcribed non-coding non-regulatory sequence as a function of the fraction of CRMs

In addition to the Manak et al. microarray data, we used the microarray data of Stolc et al. (2004) to examine whether or not the REDfly analysis CRMs are transcribed. Among the key differences between these two datasets are that (1) the Manak et al. experiments were done using total RNA while the Stolc et al. experiments used polyA-selected RNA and (2) the Manak et al. microarrays were more fully tiled, with over 15 times as many basepairs interrogated at 35 nucleotide resolution. As a result of the lower tiling density, only 149 of the CRMs have sequences represented on the Stolc et al. microarrays. Of these, 53% are transcribed (Table S5-1). We looked separately at CRMs that are promoter-proximal (within 500 bp of the TSS) and those that are more distant from the promoter. Surprisingly, the promoter-associated CRMs are significantly less likely to be transcribed (32% vs. 56%; Fisher’s Exact P< 0.04). This is not seen in the Manak et al. data (data not shown); the difference is possibly due to the higher resolution/greater coverage of the probe tiling in that dataset. Because promoter proximity presented a possible confounding factor in the Stolc et al. dataset, for the remainder of the analysis we focused on only the promoter-distal CRMs.