Collapsing dubiously resolved gene-tree branches in phylogenomic coalescent analyses

Area of Science:

• Phylogenetics
• Evolutionary Biology
• Computational Biology

Background:

• Two-step coalescent analyses assume gene tree topologies are fixed, attributing all conflict to lineage sorting.
• Empirical phylogenomic data often exhibit conflict due to gene tree estimation error, not solely lineage sorting.
• Weakly supported or arbitrarily resolved clades in gene trees contribute significantly to estimation error.

Purpose of the Study:

• Quantify the impact of different gene tree branch-collapsing methods on species tree inference.
• Evaluate the effectiveness of various branch-collapsing strategies using empirical phylogenomic datasets.
• Determine the optimal methods for collapsing gene tree branches in two-step coalescent analyses.

Main Methods:

• Applied branch-collapsing methods based on strict consensus, bootstrap thresholds, and approximate likelihood ratio test (SH-like aLRT) support.
• Analyzed seven empirical phylogenomic datasets to assess the severity of dubious gene tree resolutions.
• Quantified the effects of branch collapsing on species tree branch lengths and topology.

Main Results:

• Up to 86% of internal gene tree branches were found to be dubiously or arbitrarily resolved.
• Collapsing these branches increased inferred species tree branch lengths by up to 455%.
• Branch collapsing improved branch support and, in some cases, congruence with concatenation trees.

Conclusions:

• Severe and justified gene tree branch collapsing methods, such as strict consensus and 0% SH-like aLRT support, are recommended for two-step coalescent analyses.
• Collapsing unreliable branches can improve phylogenetic inference and reconcile coalescent-based results with concatenation approaches.
• In cases of improved congruence after collapsing, concatenation-derived resolution may be preferable, suggesting incomplete lineage sorting is not always the primary driver of conflict.