Improving phylogenetic regression under complex evolutionary models.

Area of Science:

• Evolutionary biology
• Phylogenetics
• Comparative genomics

Background:

• Phylogenetic Generalized Least Square (PGLS) is widely used for correlating species traits.
• PGLS typically assumes a homogenous model of evolutionary rates across phylogenetic trees.
• Real-world evolutionary processes often exhibit heterogeneous rates, potentially violating PGLS assumptions.

Purpose of the Study:

• To assess the robustness of PGLS under heterogeneous evolutionary models.
• To evaluate the statistical performance (Type I error and power) of PGLS when its core assumption is violated.
• To propose a method for correcting PGLS biases caused by evolutionary rate heterogeneity.

Main Methods:

• Simulated trait data under various heterogeneous evolutionary models.
• Evaluated the statistical performance of classical PGLS using Type I error and power metrics.
• Developed and tested a variance-covariance matrix transformation to adjust for heterogeneity.

Main Results:

• Classical PGLS demonstrated good statistical power but unacceptably high Type I error rates when evolutionary rates were heterogeneous.
• Overlooking rate heterogeneity in phylogenetic comparative analyses can lead to misleading conclusions due to inflated Type I errors.
• The proposed transformation effectively corrects for bias caused by model heterogeneity.

Conclusions:

• Heterogeneous rates of evolution are common, especially in large phylogenetic trees, and significantly impact PGLS.
• Ignoring evolutionary rate heterogeneity compromises the reliability of phylogenetic comparative studies.
• The developed method offers a robust solution to correct PGLS for rate heterogeneity, even without prior knowledge of the evolutionary model.