SENCA: A Multilayered Codon Model to Study the Origins and Dynamics of Codon Usage

Gene sequences are the target of evolution operating at different levels, including the nucleotide, codon, and amino acid levels. Disentangling the impact of those different levels on gene sequences requires developing a probabilistic model with three layers. Here we present SENCA (site evolution of nucleotides, codons, and amino acids), a codon substitution model that separately describes 1) nucleotide processes which apply on all sites of a sequence such as the mutational bias, 2) preferences between synonymous codons, and 3) preferences among amino acids. We argue that most synonymous substitutions are not neutral and that SENCA provides more accurate estimates of selection compared with more classical codon sequence models. We study the forces that drive the genomic content evolution, intraspecifically in the core genome of 21 prokaryotes and interspecifically for five Enterobacteria. We retrieve the existence of a universal mutational bias toward AT, and that taking into account selection on synonymous codon usage has consequences on the measurement of selection on nonsynonymous substitutions. We also confirm that codon usage bias is mostly driven by selection on preferred codons. We propose new summary statistics to measure the relative importance of the different evolutionary processes acting on sequences.