Phylospect: a spectral operator framework for detecting episodic selection in codon evolution

Area of Science:

• Molecular Evolution
• Phylogenetics
• Genomics

Background:

• Detecting episodic positive selection on phylogenetic branches is crucial in molecular evolution.
• Standard methods like PAML's branch-site test have inflated false positive rates due to multinucleotide mutations (MNMs).
• MNMs are not adequately represented by current codon substitution models.

Purpose of the Study:

• Introduce PHYLOSPECT (Phylogenetic Spectral Operator-based Codon Testing), a novel likelihood-free framework for branch-specific selection detection.
• Address the limitations of existing methods in handling MNMs.
• Provide a faster and more calibrated alternative for large-scale genomic surveys.

Main Methods:

• PHYLOSPECT utilizes an operator-based framework, estimating branch-specific codon substitution rate matrices via stochastic mapping.
• It evaluates the Nonsynonymous Operator Excess (NOE) against a parametric bootstrap null distribution.
• This approach is likelihood-free and directly models codon substitutions.

Main Results:

• Simulations show PHYLOSPECT has calibrated null behavior with a low false positive rate (0.058 at α=0.05).
• The method achieves 95% power at ω=3.0 with 3,000 codons and demonstrates increased power with alignment length and selection intensity.
• Under 3% MNM contamination, PHYLOSPECT reduced false positives by 3.3-fold compared to PAML (0.30 vs 1.00).

Conclusions:

• PHYLOSPECT offers a fast (5-10x faster than PAML) and accurate method for detecting episodic positive selection, especially in the presence of MNMs.
• It complements existing methods by detecting branch-coherent rate elevations, whereas PAML is sensitive to site-localized selection.
• PHYLOSPECT is ideal as a calibrated pre-screening tool for genome-scale evolutionary analyses.