Ancestral sequence reconstruction

The amino acid sequence of the LCA of HcydOS and HmelOS was calculated using the aaml program in PAML. The set of 17 Heliconius sequences used in the phylogenetic analysis was manually refined to minimise gaps in the multiple sequence alignment (MSA) (Heliconius_ASR_edited_aligned). This is because there are no models for insertions and deletions in PAML programs. The manually refined Heliconius sequences were aligned using Clustal Omega on the EMBL-EBI interface with default parameters [94]. A phylogenetic tree for these Heliconius sequences was generated in PhyML with 1,000 bootstraps; the best model was JTT + G + I + F. Marginal ancestral sequence reconstruction was performed using the aaml program in PAML using an empirical model and wag.DAT substitution rate matrix. To investigate intraspecific variation in the presence of these amino acid differences, we aligned sequences from whole genome sequencing of 10 individuals of H. melpomene and 10 H. cydno from Panama (S12 Table, Resequenced_mel_cyd.fa). Two sequences were generated per individual, not representing true haplotypes. We aligned amino acid sequences using Clustal Omega on the EMBL-EBI interface [94]. Alignments were visualised using BoxShade (https://embnet.vital-it.ch/software/BOX_form.html). Manually curated sequences and gene sequences from resequenced individuals are available from OSF (https://osf.io/3z9tg/).