Inferring the timescale of dengue virus evolution under realistic models of DNA substitution

Dengue virus (DENV) is the agent of the most widespread vector-borne viral disease of humans. To infer the timescale of DENV evolution with as much accuracy as possible, we compared, within a Bayesian Markov Chain Monte Carlo (MCMC) framework, estimates of phylogenetic tree length using both covarion and noncovarion models of molecular evolution, the latter also incorporating lineage-specific rate variation through a "relaxed" molecular clock. Using a data set of 32 complete genome sequences representing all four viral serotypes, we found evidence for covarion-like evolution at second codon positions in specific DENV genes, although rarely at the level of complete gene or genomes. Further, the covarion model had little effect on estimates of tree length and hence time to the Most Recent Common Ancestor (MRCA). We conclude that although covarion models can improve descriptions of the dynamics of amino acid substitution, they have little effect on estimates of the timescale of viral evolution, which in the case of DENV covers a period of no more than 2000 years.