Gene-wise resampling outperforms site-wise resampling in phylogenetic coalescence analyses

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Coalescent analyses are crucial for inferring species trees from genomic data.
• Bootstrap resampling is commonly used to assess branch support in phylogenetic analyses.
• The choice between site-wise and gene-wise bootstrapping can impact support values.

Purpose of the Study:

• To compare the reliability of site-wise versus gene-wise bootstrapping in coalescent-based phylogenetics.
• To determine which bootstrapping method provides more conservative branch support for species trees.
• To evaluate the impact of bootstrapping strategies on gene tree estimation error.

Main Methods:

• Applied site-wise and gene-wise bootstrap analyses to three empirical phylogenomic datasets.
• Utilized four coalescent methods: ASTRAL, MP-EST, NJst, and STAR.
• Developed and used a script for gene-wise resampling (bootstrap and jackknife).

Main Results:

• Site-wise bootstrapping generally led to increased gene tree conflicts and inflated support for incorrect clades.
• Gene-wise bootstrapping consistently provided more conservative branch support across all tested coalescent methods and datasets.
• Gene-wise resampling outperformed site-wise resampling in reducing gene tree estimation error.

Conclusions:

• Site-wise bootstrapping is not a reliable method for conservative branch support in summary coalescent analyses.
• Gene-wise resampling is recommended over site-wise or combined approaches for robust phylogenetic inference, particularly with large numbers of genes.
• A gene-wise resampling script is provided to facilitate more accurate phylogenetic analyses.