Miyasaka's model of substitution solid solution is assumed for the crystalline complex.
We therefore determined to use the WAG model of substitution for all loci, optimizing model parameters separately for each locus; see Supplementary Fig. 1, panel (iii).
The model of substitution was based on the suggestion by MODELTEST (http://darwin.uvigo.es/software/modeltest.html).html
Coding sequences were partitioned by codon and the analyses were run using a GTR+I+γ model of substitution selected using jModel test [66].
Amino-acid analysis was based on the POISSON model and nucleotide analysis adopted a codon-based GTR+G+I model of substitution.
Maximum likelihood (ML) analyses on the moth and plant matrices were performed under the GTR + G model of substitution using RAxML [34].
Phylogenetic trees were reconstructed using MEGA3 [64], from aligned cDNA sequences using the neighbor-joining method with the Kimura-2 parameter model of substitution [65].
To estimate past population sizes we used BEAST 1.4.8 [47] to construct a Bayesian skyline plot (employing the Bayesian MCMC coalescent method, a GTR+G+Γ model of substitution [MODELTEST 3.7 [48]], and a strict clock).
The site by site rate variation was set to a gamma distribution (rates = gamma) for all the Bayesian trees and a General Time-Reversible (GTR) (nst = 6) model of substitution was chosen.
We used the Jones-Taylor-Thornton (JTT) model of substitution, with the frequencies of amino acids being estimated from the data set, and rate heterogeneity across sites modeled in eight rate categories.
ML and BI analyses of the amino acid alignment (i.e., alignment 4) was analyzed under a WAG model of substitution considering corrections for site-to-site rate variation (gamma) with eight categories of rate variation and proportion of invariable sites.
