The guide-tree analysis was performed on a set of 48 bird genomes previously published29 (link). To reduce the amount of alignment work required, we subset these genomes down to the size of only a single chromosome, chicken chromosome 1 (by removing any contig or scaffold that had less than 20% of its sequence alignable to chicken chromosome 1). We used Progressive Cactus commit 36304707 for all alignments in this analysis.
The Prum and Jarvis topologies were adapted from Prum et al.28 (link) and Jarvis et al.29 (link), respectively. The ‘permuted’ topology was generated starting from the Jarvis topology, via three randomly chosen subtree-prune-regraft operations followed by three random nearest-neighbour-interchange operations. Each of these three topologies had branch-length estimates performed using phyloFit from the PHAST package (https://github.com/CshlSiepelLab/phast , commit 52e8de9) based on fourfold-degenerate sites of BUSCO orthologues. Finally, the ‘consensus’ tree was produced as a strict consensus of the Jarvis and Prum trees (collapsing all groupings that were not the same in both trees) using the ape::consensus method from the APE R package44 (link). The branch-lengths for this tree were generated from the fitted branch lengths for the two input trees, using the consensus.edges function of the phytools R package45 . The four final trees that were used in the four Progressive Cactus alignments are shown in Supplementary Fig. 1 , and available in supplementary data in Newick format.
We further focused on the alignments with guide trees based on Jarvis29 (link) and Prum28 (link) (Supplementary Fig.3 ) to establish what alignment differences resulted from different phylogenetic hypotheses. Supplementary Fig. 2 shows a refinement of the overall alignment-to-alignment F1 scores shown in Extended Data Table 2 , showing the F1 scores for each species pair between the Jarvis- and Prum-based alignments. Each pair of species has an F1 score between Jarvis- and Prum-based alignments of at least 0.955.
The Prum and Jarvis topologies were adapted from Prum et al.28 (link) and Jarvis et al.29 (link), respectively. The ‘permuted’ topology was generated starting from the Jarvis topology, via three randomly chosen subtree-prune-regraft operations followed by three random nearest-neighbour-interchange operations. Each of these three topologies had branch-length estimates performed using phyloFit from the PHAST package (
We further focused on the alignments with guide trees based on Jarvis29 (link) and Prum28 (link) (Supplementary Fig.
Free full text: Click here
