The Limits of the Constant-rate Birth-Death Prior for Phylogenetic Tree Topology Inference

Area of Science:

• Evolutionary biology
• Phylogenetics
• Computational biology

Background:

• Birth-death models are crucial for inferring evolutionary timescales.
• Previous studies indicate discrepancies between constant-rate birth-death (crBD) model expectations and empirical phylogenetic trees, particularly in tree imbalance.
• A comprehensive understanding of these differences and the model's practical limitations is needed.

Purpose of the Study:

• To quantify the topological differences between crBD model expectations and empirically inferred phylogenies.
• To assess the practical limits and robustness of the crBD model in phylogenetic inference.
• To investigate the impact of crBD priors on tree topology inference under varying evolutionary rates and tree imbalance.

Main Methods:

• Comparison of crBD model expectations against a dataset of 1189 empirical trees using various topology indices.
• Meta-analysis of studies employing Bayesian methods with crBD priors versus non-crBD priors and non-Bayesian methods.
• Simulation of sequence data on highly imbalanced empirical tree topologies under different evolutionary rates, followed by tree re-inference using maximum likelihood and Bayesian crBD models.

Main Results:

• crBD model trees frequently exhibit topological differences compared to empirical trees.
• No significant differences in tree topology inferences were found between Bayesian crBD priors and other methods in a meta-analysis.
• Under low substitution rates, the crBD prior tends to produce overly balanced trees, a tendency that diminishes at higher substitution rates.

Conclusions:

• The crBD prior demonstrates general robustness across diverse phylogenetic inference scenarios.
• Empirically observed phylogenetic imbalance is unlikely under the crBD model, suggesting potential systematic bias in datasets with limited information.
• The crBD model's limitations are most apparent in highly imbalanced trees and low substitution rate scenarios.