Complex Models of Sequence Evolution Require Accurate Estimators as Exemplified with the Invariable Site Plus Gamma...

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• The invariable site plus Gamma (I+Γ) model is standard for modeling rate heterogeneity in phylogenetic analyses.
• Identifiability of the continuous I+Γ model is proven, enhancing its credibility.
• The discrete I+Γ model, commonly used in software, lacks proven identifiability and parameter estimation is debated.

Purpose of the Study:

• To investigate the reliability of parameter estimation for the discrete I+Γ model in phylogenetic software.
• To address the problematic correlation between invariable sites and slow evolutionary rates.
• To develop an improved optimization heuristic for accurate parameter estimation.

Main Methods:

• Evaluation of parameter estimation accuracy in common phylogenetic software (PhyML, RAxML, IQ-TREE, MrBayes).
• Analysis of optimization heuristics' performance with the discrete I+Γ model.
• Development and testing of a novel optimization heuristic.

Main Results:

• Existing phylogenetic software optimization heuristics frequently fail to reliably estimate key I+Γ model parameters (shape, proportion of invariable sites, tree length).
• A significant correlation exists between the proportion of invariable sites and sites with slow evolutionary rates, complicating estimation.
• The proposed improved optimization heuristic demonstrates accurate estimation of these critical parameters.

Conclusions:

• Reliable estimation of parameters for complex models of sequence evolution is as vital as tree search methods.
• The developed heuristic offers a more accurate approach to estimating parameters for the discrete I+Γ model.
• Further development and verification of estimation methods are crucial for advancing phylogenetic reconstruction.