Accounting for background nucleotide composition when measuring codon usage bias.

The phenomenon of codon usage bias has been important in the study of evolution because it provides examples of weak selection working at the molecular level. During the last two decades, evidence has accumulated that some examples of codon usage bias are driven by selection, particularly for species of fungi (e.g., Bennetzen and Hall 1982; Ikemura 1985), bacteria (e.g., Ikemura 1981; Sharp and Li 1987), and insects (e.g., Akashi 1997; Moriyama and Powell 1997). This connection between codon usage bias and selection has been important in stimulating the development of alternatives to strictly neutral theories of molecular evolution (Ohta and Gillespie 1996; Kreitman and Antezana 2000). Uncertainty persists, however, regarding the phylogenetic distribution of codon usage bias (such as whether selection-based codon usage bias is present in mammals; e.g., Karlin and Mrazek 1996; Iida and Akashi 2000; Sueoka and Kawanishi 2000; Smith and EyreWalker 2001a; Urrutia and Hurst 2001). Questions also remain as to what models of selection underlie codon preferences (Kreitman and Antezana 2000), and specifically whether the presence of suboptimal codons is the result of mutation and drift, variation in selection pressure across sites, or antagonistic selection pressures (Smith and Eyre-Walker 2001b). For investigating certain questions regarding the evolution of codon usage bias, it is useful to have a summary statistic describing the pattern of codon usage across all amino acids. Many summary statistics have already been developed to describe the patterns of codon usage. They can be divided roughly into two classes (Comeron and Aguade 1998). One class summarizes the usage of certain preferred codons, and the other compares every codon’s usage to a null distribution (typically uniform usage of synonymous codons). The former class of methods has the disadvantage that it requires a prior knowledge of the preferred codons. With summary statistics one can explore general patterns, such as the relationship of codon usage bias to recombination rate, gene length, or synonymous substitution rate. Observing broad patterns such as these has already provided insight into the evolutionary dynamics of codon usage bias (Kliman and Hey 1993; Akashi and Eyre-Walker 1998; Kreitman and Comeron 1999).