Recodon: coalescent simulation of coding DNA sequences with recombination, migration and demography

Abstract

Background: Coalescent simulations have proven very useful in many population genetics studies. In order to arrive to meaningful conclusions, it is important that these simulations resemble the process of molecular evolution as much as possible. To date, no single coalescent program is able to simulate codon sequences sampled from populations with recombination, migration and growth.

Results: We introduce a new coalescent program, called Recodon, which is able to simulate samples of coding DNA sequences under complex scenarios in which several evolutionary forces can interact simultaneously (namely, recombination, migration and demography). The basic codon model implemented is an extension to the general time-reversible model of nucleotide substitution with a proportion of invariable sites and among-site rate variation. In addition, the program implements non-reversible processes and mixtures of different codon models.

Conclusion: Recodon is a flexible tool for the simulation of coding DNA sequences under realistic evolutionary models. These simulations can be used to build parameter distributions for testing evolutionary hypotheses using experimental data. Recodon is written in C, can run in parallel, and is freely available from http://darwin.uvigo.es/.

Figure 1

Effect of population structure on the estimation of synonymous and nonsynonymous divergence. Nine different scenarios were simulated, combining three migration rates (m = 0.0002, 0.01 and ∞ (= one deme)) and three dN/dS ratios (dashed line = 0.1, solid line = 1, dotted line = 10). For each scenario, 500 alignments with 10 sequences 333 codons long, were simulated. In all cases, the mutation rate was 5.4 ∞ 10^-5, the transition/transversion ratio was 1.0, and the effective population size was 1000. Mean synonymous divergence per synonymous site (dS), nonsynonymous divergence per nonsynonymous site (dN), and their ratio (dN/dS) were estimated according to Nei and Gojobori [32] with a modified version of SNAP [33]. Error bars indicated approximate 95% confidence intervals (± s.e. ∞ 1.96).