Figure 4

A. Conserved selection for strong / weak folding related signals cannot be explained basing only on dinucleotide composition. MFE related signal conservation with respect to evolutionary-constrained randomization model is shown in positions that overlap with MFE conserved signals with respect to dinucleotide-constrained randomization model along the coding regions of 4 DENV serotypes for strong (red) and weak (green) folding. As many as 60%, 52%, 49%, 34% of positions with conserved signals related to strong folding (red) and 62%, 58%, 43%, 44% of positions possessing weak folding signal conservation (green) (for serotypes 1 - 4 correspondingly) overlapped with MFE conserved signals identified with respect to dinucleotide-constrained randomization model, and this overlap was not likely to appear in random (p-value < 0.001; no overlap was observed in 1000 randomized variants). B. The regions with significantly conserved strong / weak folding signal cannot be explained based only on sequence conservation. A low / insignificant Spearman correlation between conservation levels of MFE related signals (x-axis) and the nucleotide / synonymous variability (y-axis) in the corresponding genomic intervals is demonstrated. The least square linear approximation of the data points is plotted in blue.