In order to verify this hypothesis, we compared the tissue specificity between the ubiquitously expressed genes (expression breadth ≥ 14) and the intermediately expressed genes (expression breadth = 8 or 9) with similar expression level.
For the comparison of gene expression breadth, expressed miRNA and non-miRNA target genes were first extracted from the dataset.
For each gene, we then summed up the number of over expressed tissues to compute tissue expression breadth.
The difference in the gene expression breadth between gbM and unmethylated genes was due to a higher proportion of the gbM genes (82.4 vs. 72% unmethylated genes) having a maximal expression breadth.
Additionally, genes with a high expression breadth had fewer reads mapping to their introns than genes with low expression breadth (Table 3), indicating that selection could act on reducing intron retention in potentially key genes.
Finally, genes with different expression breadth show different rates of protein evolution (Slotte et al. 2011) and thus expression breadth could also affect variation in the observed amount of expression noise of a gene.
Identically, genes with a high expression breadth could primarily be key genes, like housekeeping genes, and could be under selection to minimise their expression noise.
Average intron density is correlated to expression breadth in both humans [13] and Arabidopsis.
Second, PCI- genes have a higher evolutionary rate and lower expression breadth than PCI+ genes.
Non coding gene sequences, showed different patterns between expression breadth and level: introns and UTRs were positively correlated to expression breadth and negatively or not even associated to expression level.
As expected intron density was positively correlated to expression breadth for both oligo-array and MPSS experiments (Table 1).
