(A) Examples of local synonymous rate variation in alignments of 29 placental mammals for short nine-codon windows within the open reading frames (ORFs) of three known human protein-coding genes—ALDH2, BMP4, and GRIA2—with brackets denoting starting codon position within each ORF of shown alignment. (Bright green) Synonymous substitutions with respect to the inferred ancestral sequence; (dark green) conservative amino acid substitutions; (red) other nonsynonymous substitutions. The estimated parameter λsome denotes the rate of synonymous substitution within these selected windows relative to genome-wide averages. For example, the nine-codon window starting at codon 88 of the BMP4 ORF shows λsome = 0.5, corresponding to an estimated synonymous substitution rate 50% below the genome average. (B) Variation in the estimated synonymous rate at different positions with respect to exon boundaries and translation start and stop, across all CCDS ORFs. For each class of regions, box-and-whisker plots show the observed distribution of λsome, including the median (middle horizontal bars), middle 50% range (boxes), extreme values (whiskers), and whether medians differ with high statistical confidence (nonoverlapping notches between two boxes). Estimated synonymous rates tend to be significantly reduced at the 5′ and 3′ ends of exons, and dramatically reduced in alternatively spliced exons, likely reflecting widespread splicing regulatory elements embedded within protein-coding regions.
