Figure 7.

Human exons are optimized for low indel propensity. The fraction of positions with indel propensity score above 1 (indel-susceptible loci) was computed across all exons in the human genome. The graphs show the ratio between these fractions and the fraction of such loci in exons that were randomized by shuffling (preserving GC content) synonymous codons. We see a clear preference of synonymous codons for low indel propensity, except for indels of size 3, which do not cause frame shifts.