Deficit of 3n IESs in coding sequences.

For the calculation of χ2, the observed numbers of IESs of length 3n and non-3n inserted in coding sequences are compared to the distribution found for IESs inserted in non-coding sequences under the null hypothesis that IES length is not under constraints related to translation. The null hypothesis is rejected only for those IESs inserted in coding sequences that do not contain a stop codon in frame with the upstream ORF (Sample “Coding stopless”). Microarray experiments [38] were used to group the IESs according to the expression level of the genes in which they are inserted. “Q1” designates IESs in exons of the 25% least expressed genes and “Q4” designates IESs in the exons of the 25% most expressed genes, those subject to the strongest selective pressure. The bias against 3n IESs is stronger in the Q4 sample than in the Q1 sample.A more detailed analysis of the modulo 3 length distribution for IESs in coding and non-coding sequences, for each peak of the 10 bp periodic size distribution, is provided in Table S5.