Weighted bootstrapping: a correction method for assessing the robustness of phylogenetic trees

Area of Science:

• Phylogenetics and evolutionary biology
• Statistical methods in bioinformatics

Background:

• Non-parametric bootstrapping is standard for assessing phylogenetic tree confidence.
• Traditional methods do not weigh trees by their quality, potentially skewing confidence scores.
• Tree quality can be measured by fit scores like least-squares (LS) or parsimony.

Purpose of the Study:

• To introduce and evaluate weighted bootstrapping for phylogenetic reconstruction.
• To assess the impact of weighting individual resampled trees on overall bootstrap support.
• To compare novel weighting strategies against traditional unweighted bootstrapping.

Main Methods:

• Applying weighted bootstrapping to phylogenetic inference.
• Using tree quality metrics, specifically least-squares (LS) and secondary bootstrap score (SBS), for weighting.
• Conducting simulation studies to compare five weighting strategies: LS-based, SBS-based, LS-based with data normalization, SBS-based with data normalization, and traditional unweighted bootstrapping.

Main Results:

• SBS-based bootstrapping with data normalization generally yielded higher bootstrap scores and robustness.
• The normalized SBS approach outperformed traditional bootstrapping and other weighting strategies.
• Other strategies showed similar performance, indicating stability of traditional methods.

Conclusions:

• Normalized SBS-based bootstrapping is a promising method for robust phylogenetic confidence assessment.
• This method is particularly advantageous when dealing with noisy or heavily modified sequence data.
• The study highlights the importance of incorporating tree quality into bootstrap calculations for more reliable phylogenies.