Phylogenetic relationships
We built a multiple sequence alignment (MSA) for each 1:1 orthologs with PRANK v.100802 (Löytynoja, 2014), using both coding sequences (CDS) (-codon -noxml -notree -F) and amino acid sequences (-noxml -notree -F). Individual alignments were concatenated with catfasta2phyml v1.1.0 (https://github.com/nylander/catfasta2phyml) to create a CDS supermatrix and an amino acid supermatrix. Only locus with data for all the twelve species have been considered. Fourfold degenerate sites (4D) for the CDS supermatrix were extracted with MEGA X (Kumar, Stecher, Li, Knyaz, & Tamura, 2018). We performed unpartitioned maximum likelihood (ML) phylogenetic analyses using IQ-TREE 1.6.12 (Nguyen, Schmidt, Von Haeseler, & Minh, 2015) (-bb 1000) both for 4D and amino acid supermatrices. Optimal models of sequence evolution were obtained with ModelFinder (Kalyaanamoorthy, Minh, Wong, Von Haeseler, & Jermiin, 2017) according to Bayesian information criterion (BIC), and the resulting best-fit models were GTR+F+R2 for 4D and HIVb+F+R3 for amino acid supermatrix. Node support was obtained with Ultrafast Bootstrap (Hoang, Chernomor, Von Haeseler, Minh, & Vinh, 2018).
To explicitly account for incomplete lineage sorting (ILS) under the Multispecies Coalescent Model (MSC), we inferred the species tree using the summary coalescent approach, as implemented in ASTRAL-III 5.6.3 (C. Zhang, Rabiee, Sayyari, & Mirarab, 2018). We first obtained all gene trees (for each 1:1 orthologous genes) using IQ-TREE 1.6.12, and inferred the species tree and its normalized score (from both CDS and amino acid gene trees) using ASTRAL-III.
We generated an ultrametric tree with r8s v.1.81 (Sanderson, 2003) using the 4D supermatrix ML tree. We used four calibration points (in myr): root (max_age=84 min_age=73, (Braun et al., 2011), most recent common ancestor (MRCA) of Spheniscidae (min_age=12.6, (Subramanian, Beans-Picón, Swaminathan, Millar, & Lambert, 2013)), MRCA Procellariiformes (min_age=49, Claramunt & Cracraft, 2015), MRCA Procellariidae (min_age=14, (Prum et al., 2015)), retrieved from TimeTree (Kumar, Stecher, Suleski, & Hedges, 2017), and references therein). We used the penalized likelihood (PL) method and the Truncated Newton (TN) algorithm, smoothing parameter was set to 100.