The Multispecies Coalescent Model Outperforms Concatenation Across Diverse Phylogenomic Data Sets

Area of Science:

• Phylogenomics and evolutionary biology
• Statistical modeling in bioinformatics
• Computational phylogenetics

Background:

• Debates persist regarding concatenation versus coalescent models in phylogenomic data analysis.
• Statistical frameworks for model comparison and validation are essential for resolving these debates.
• Previous studies have highlighted potential issues with standard phylogenetic modeling approaches.

Purpose of the Study:

• To statistically evaluate and compare the adequacy of substitution, concatenation, and multispecies coalescent (MSC) models.
• To assess the widespread applicability and limitations of these models across diverse phylogenomic datasets.
• To provide insights into the reliability of gene tree congruence and its implications for species tree estimation.

Main Methods:

• Application and development of statistical tests for model evaluation.
• Analysis of 47 phylogenomic datasets spanning the tree of life.
• Utilized logistic regression and Bayesian model validation/comparison techniques.

Main Results:

• Substitution and concatenation models showed widespread poor fit, rejected by 44% and 38% of loci, respectively.
• GC content and informative site proportions negatively correlated with substitution model fit.
• The multispecies coalescent (MSC) model was favored over concatenation, especially for larger datasets (>10 loci).

Conclusions:

• The concatenation assumption of congruent gene trees is frequently violated in phylogenomic data.
• The MSC model provides a more robust framework for phylogenomic inference compared to concatenation.
• Model validation and comparison are critical for advancing phylogenomic analyses and improving computational tools.