Structure-informed evolutionarily constrained regions. (A) Plot showing the fraction of bases identified as evolutionarily constrained at various FDRs (21) by the DNA sequence-based algorithm binCons (blue line) and the structure-informed algorithm Chai (red line). (B) Venn diagram showing nucleotides identified by each algorithm [Chai (red) and binCons (blue)] at 5% FDR (gray vertical line in A). Chai detects nearly the same set of bases as binCons identifies (intersection), plus a nearly equal number of additional bases (Chai-only). (C) Bar graph showing the fraction of different types of genomic regions that overlap binCons- (blue), Chai- (red), and Chai-only-detected (white) constrained elements. Black points are the mean of a null distribution constructed with the GSC method (21); error bars represent 95% confidence intervals.

Luciferase-based reporter activity of 12 regions containing Chai-detected elements (21). These regions overlapped predicted enhancer regions. Plotted for each element is the luciferase activity relative to the median activity from 100 random control constructs (y-axis; see Fig. S5). Error bars represent one standard deviation from the mean of four experimental replicates, and asterisks denote p ≤ 0.05 (Wilcoxon rank sum test).

DNA topography vs. nucleotide sequence. (A–C) Hydroxyl radical cleavage patterns (referred to here as the structural profile) and corresponding color-matched sequence alignments (below each panel). Asterisks indicate identical columns in the sequence alignment. (D–G) Single-base substitutions have a range of effects on local DNA structure. For all possible 11-bp nucleotide sequences, we computationally changed the middle position to each possible base and measured its effect on the structural profile. These changes were quantitatively measured by calculating the mean Euclidean distance (referred to here as the average structural change) (21), where low values indicate similar structure profiles and large values indicate different structure profiles. (G) The distribution of average structure changes observed for all 11-bp sequences. Arrows indicate what we classify as low, moderate, or high average structure changes, with representative 11-bp sequences for each shown in (D), (E), and (F), respectively. For (D–F), the y-axis indicates the structural profile at each nucleotide position in the 11-bp sequence (x-axis). The structural profile for each sequence containing a different base in the middle position (noted by an ‘N’) is plotted in a different color. The structural profile patterns are increasingly different in changing from low to high structural change [(E–G)].

The distribution of DNA structural changes for phenotype-associated variants (red line) and neutral variants (black line) (21) with the real structure prediction algorithm (A) and a randomized version (B). (A) The phenotype-associated distribution was shifted significantly to the right (p < 3 × 10⁻⁴; Wilcoxon rank sum test) and contains an additional high structure change peak relative to the neutral distribution. (B) No significant shift (p > 0.05; Wilcoxon rank sum test) was observed with a randomized version of the structure prediction algorithm.