Distance-Based Phylogenetic Placement with Statistical Support.

Area of Science:

• Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• Phylogenetic identification of unknown sequences is vital in ecological studies.
• Incomplete and noisy data necessitate uncertainty measures for phylogenetic placements.
• Standard likelihood-based methods offer uncertainty quantification, but newer distance-based methods lack this.

Purpose of the Study:

• To adapt and evaluate uncertainty quantification methods for distance-based phylogenetic placement.
• To compare the accuracy of different sampling strategies for measuring phylogenetic placement support.
• To optimize the efficiency of bootstrapping for distance-based phylogenetic analyses.

Main Methods:

• Applied parametric and nonparametric sampling methods to measure support for distance-based phylogenetic placements.
• Developed a linear algebraic formulation for efficient bootstrapping, achieving up to 30x speedup.
• Integrated optimized bootstrapping into the APPLES software for distance-based placement.

Main Results:

• Nonparametric bootstrapping demonstrated higher accuracy for phylogenetic placement support compared to other tested methods.
• The optimized bootstrapping method significantly increased computational efficiency.
• Maximum likelihood (ML) support values were more accurate for fragmentary queries, while distance-based methods excelled with full-length and multi-gene datasets.

Conclusions:

• Nonparametric bootstrapping effectively quantifies uncertainty in distance-based phylogenetic placements.
• Optimized bootstrapping enhances the speed and accuracy of distance-based phylogenetic tools.
• This work bridges the gap in uncertainty quantification, promoting wider adoption of distance-based phylogenetic placement methods.