Fig. 1

Common regions of folding bias (∆LFE) are present across the tree of life, but are not universal. There is correlation between the strengths of these regions in different species, indicating there are factors influencing the bias throughout the coding sequence. a Summary of profile features with the fraction of species in which each feature appears in each domain (based on model 1 rules; see “Analysis” under the “Methods” section for details). The results based on the less restrictive model 2 rules (with weaker ∆LFE near the CDS edges not required to be positive; see “Analysis” under the “Methods” section) are shown in bright blue below each bar. References shown here are based on comparison to randomized sequences (i.e., equivalent to ∆LFE). b Scheme illustrating profile features reported separately in previous studies within the CDS, showing features [A]–[D] from a. c Observed distribution of ∆LFE profile values at different positions relative to CDS start (left) and end (right). d The distances (in nt) from the start codon where ∆LFE transitions from positive to negative, for species belonging to different domain. The lengths of the initial weak folding region range up to 150 nt in some bacteria. e Spearman’s correlations between mean ∆LFE profile values in regions [A], [C], and [D]. White dots indicate significant correlation (p value < 0.01)