Declining transition/transversion ratios through time reveal limitations to the accuracy of nucleotide substitution models

BACKGROUND: Genetic analyses of DNA sequences make use of an increasingly complex set of nucleotide substitution models to estimate the divergence between gene sequences. However, there is currently no way to assess the validity of nucleotide substitution models over short time-scales and with limited mutational accumulation. RESULTS: We show that quantifying the decline in the ratio of transitions to transversions (ti/tv) over time provides an in-built measure of mutational saturation and hence of substitution model accuracy. We tested this through detailed phylogenetic analyses of 10 representative virus data sets comprising recently sampled and closely related sequences. In the majority of cases our estimates of ti/tv decrease with time, even under sophisticated time-reversible models of nucleotide substitution. This indicates that high levels of saturation are attained extremely rapidly in viruses, sometimes within decades. In contrast, we did not find any temporal patterns in selection pressures or CG-content over these short time-frames. To validate the temporal trend of ti/tv across a broader taxonomic range, we analyzed a set of 76 different viruses. Again, the estimate of ti/tv scaled negatively with evolutionary time, a trend that was more pronounced for rapidly-evolving RNA viruses than slowly-evolving DNA viruses. CONCLUSIONS: Our study shows that commonly used substitution models can underestimate the number of substitutions among closely related sequences, such that the time-scale of viral evolution and emergence may be systematically underestimated. In turn, estimates of ti/tv provide an effective internal control of substitution model performance in viruses because of their high sensitivity to mutational saturation. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1186/s12862-015-0312-6) contains supplementary material, which is available to authorized users.