“…Our tree is the maximum-clade credibility tree summarizing a posterior distribution of trees inferred by BEAST 2 (14). Genomic sites were partitioned into the 3 codon positions, plus noncoding, as described (3) For each partition, we modeled evolution with an HKY (Hasegawa-Kishino-Yano) substitution model with log-normal(μ = 1, σ = 1.25) prior on κ, frequencies estimated with Dirichlet (1,1,1,1) prior, and relative substitution rates with Dirichlet (1,1,1,1). We used a strict clock model with log-normal(μ = −7, σ = 1.25) prior on mean clock rate, and for the tree prior we used a Bayesian skyline model (15) with Markov chain distribution on population sizes and log-normal(μ = 0, σ = 2) on the first population size.…”