A. Comparison between the mRNA folding energy of randomized windows that maintain the window amino acid content (Methods) with ATGs and without ATGs.

<p>The folding energy between the two sets of windows is not significantly different at the 5′UTR; at the beginning of the ORF windows with more ATGs tend to have <i>weaker</i> mRNA folding. Positions with a significant p-value (paired t-test) are marked by a red dot. Thus, the fact that there is selection for weak mRNA folding at the beginning of genes cannot explain why there are less ATGs at the beginning of genes. B. Genomic profiles of number of ATGs in the first 14 codons following the START ATG, in comparison to 20 randomized genomes that maintain the <i>amino acid order of each gene</i>, <i>the genomic GC content and codon usage bias at the first 40 codons</i>, in frame 1 and 2 in <i>S. cerevisiae</i> (<a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1003136#s4" target="_blank">Methods</a>). Red dots – denotes positions where the random genome includes significantly more ATGs than the real genome; green dots – denotes cases in which the random genome includes significantly less ATGs than the real one. C. The number of ATG profiles in the three frames in <i>S. cerevisiae</i> and <i>S. pombe</i> for highly and lowly expressed genes (top/bottom 15% ribosomal load). The fact that there are less ATGs at the <i>beginning of the ORF</i> is more striking in the case of the highly expressed genes. P-values for comparisons of the number of ATGs in the 5′UTR and ORF, and in each of the frames separately in the highly and lowly expressed genes, appear in the figure and are specifically significant for frame 1 and 2 (<a href="http://www.ploscompbiol.org/article/info:doi/10.1371/journal.pcbi.1003136#s4" target="_blank">Methods</a>; when considering all the frames together the 5′UTR/ORF p-values are 1.07·10⁻¹² and 3.34·10⁻²⁷ respectively in <i>S. cerevisiae</i>, and 8.75·10⁻²¹ and 8.6·10⁻²⁷ respectively in <i>S. pombe</i>).