Palindromes, widespread regulatory elements in the promoters of protein coding genes [27], are less frequent around TSSs of lncRNA genes (Figure 1d). CpG islands (CGIs) are known to overlap with about two-thirds of protein-coding gene promoters [28]. Although CGIs are also hosting numerous non-coding transcripts [29], [30], an observation independent of the method of CGI detection [31], such ncRNAs are most likely short and unprocessed. On the contrary, we find that lncRNA promoters quite rarely overlap with CGIs (Figure 1e). LncRNA exons and splice junctions have been reported as enriched in repetitive elements [32]. We show that repetitive elements are also enriched at lncRNA promoters (Figure 1f, Figure S1). DNA sequence properties of non-zero similarly expressed protein-coding and lncRNA genes show feature patterns similar to those of the whole promoter sets without considering any expression levels (Figure S1).
