Probing instructions for expression regulation in gene nucleotide compositions
Fig 4
Contribution of additional genomic regions.
Genomic regions were ranked according to their contribution in predicting gene expression. First, all regions were tested separately. Introns yielded the highest Spearman correlation between observed and predicted expressions (in a cross-validation procedure) and was selected as the ‘first’ seed region. Second, each region not already in the model was added separately. 5’UTR in association with introns yielded the best correlation and was therefore selected as the ‘second’ region. Third, the procedure was repeated till all regions were included in the model. The contribution of each region is then visualized starting from the most important (left) to the less important (right). Note that the distance between the second TSS and the first ATG is > 2000 bp for only 189 genes implying that 5’UTR and DD regions overlap. The correlations computed at each steps are indicated in (S2 Table). ns, non significant.