A comparison of ancestral state reconstruction methods for quantitative characters

Area of Science:

• Evolutionary Biology
• Phylogenetics
• Quantitative Genetics

Background:

• Ancestral state reconstruction is crucial for understanding trait evolution.
• Numerous methods exist for quantitative characters, leading to confusion in selection.
• Comparing method performance under various evolutionary models is necessary.

Purpose of the Study:

• To compare the accuracy of seven distinct ancestral state reconstruction methods.
• To assess method performance under Brownian motion with and without directional or stabilizing selection.
• To develop an efficient evaluation scheme for reconstruction accuracy.

Main Methods:

• Maximum Likelihood (ML)
• Restricted Maximum Likelihood (REML)
• Generalized Least Squares (GLS) under Brownian, Brownian-with-trend, and Ornstein-Uhlenbeck models
• Phylogenetic Independent Contrasts (PIC)
• Squared Parsimony

Main Results:

• ML, REML, and GLS under Brownian models yield identical ancestral states but differ in uncertainty estimation.
• Accurate reconstruction requires detecting and accounting for evolutionary trends.
• All tested methods perform well under stabilizing selection.
• Maximum likelihood methods demonstrate superior accuracy in the absence of trends or stabilizing selection.

Conclusions:

• The choice of ancestral state reconstruction method significantly impacts accuracy.
• Detecting evolutionary trends is critical for appropriate method selection.
• Maximum likelihood approaches offer robust performance for quantitative trait reconstruction.