2.5 Phylogenomic analysis
We reconstructed phylogenetic trees using one representative sample from each of the six A. cantabrica populations and all other taxa. Some genome skimming data available online (Suppl. Table 2) were added for reconstructing the plastid phylogeny.
Loci sequences were aligned with MAFFT (–auto; Katoh & Standley, 2013), and then the alignments were trimmed with trimAl (-automated1; Capella-Gutiérrez et al. , 2009). For both Angiosperms353 loci and the 125 recovered plastid fragments, we inferred nuclear and plastid phylogenetic trees using both coalescent and concatenated maximum likelihood (ML) approaches. In the coalescent approach, we inferred single-locus phylogenetic trees from each trimmed alignment using IQ-TREE (Minh et al. , 2020) with 1000 ultrafast bootstrap replicates (-bb 1000; Hoang et al. , 2018), and branches with less than 10% bootstrap support were collapsed with Newick utilities (Junier & Zdobnov, 2010). We then used ASTRAL-III (Zhang et al. , 2018) to infer the species tree (hereafter, ASTRAL tree), applying the ”-t 3” flag to annotate local posterior probabilities (LPP) for each node.
In the concatenated ML approach, all trimmed alignments were concatenated with FASconCAT-G (Kück & Meusemann, 2010). The best-fit model inferred by IQ-TREE (-m MFP) was applied in RAxML-NG (Kozlovet al. , 2019) to infer the species tree using the concatenated partitioned matrix with 1000 bootstrap replicates (–tree pars{20} –bs-trees 1000; hereafter, RAxML tree). Additionally, we implemented a greedy strategy (Lanfear et al. , 2012) with the relaxed hierarchical clustering algorithm (Lanfear et al. , 2014) to select the best partition model, which was applied in IQ-TREE to infer the species tree with 1000 SH-like approximate likelihood ratio test replicates (-alrt 1000; Guindon et al. , 2010) and 1000 ultrafast bootstrap replicates (hereafter, IQ-partition tree). We visualized the phylogenetic trees using Dendroscope (Huson & Scornavacca, 2012) and FigTree (available at https://github.com/rambaut/figtree).
To investigate potential phylogenetic conflicts and signs of reticulate evolution, we used SplitsTree4 (Huson & Bryant, 2006) to create a split network based on the Neighbor-Joining algorithm with the Angiosperms353 data. In the resulting network, we masked specific samples to retain only those within the A. cantabrica, A. adfinis and /halleri clades for focused analysis.