When Complex Models Fit the Wrong Mechanistic Complexity in Phylogenomic Analysis.

Area of Science:

• Phylogenetics and Evolutionary Biology
• Computational Biology
• Genomics

Background:

• Substitution model selection is crucial for phylogenetic inference, typically assuming a single tree topology for all genomic sites.
• Phylogenomic data often exhibit gene-tree discordance due to factors like incomplete lineage sorting, violating the shared-genealogy assumption.

Purpose of the Study:

• To examine how unmodeled genealogical heterogeneity confounds substitution model selection in phylogenomics.
• To highlight the impact of this confounding on phylogenetic estimation, divergence-time inference, and parameter interpretation.
• To advocate for integrating genealogical heterogeneity into model selection and adequacy assessment.

Main Methods:

• Review and synthesis of recent findings on genealogical heterogeneity and substitution model selection.
• Conceptual examination of the consequences of violating shared-genealogy assumptions.
• Discussion of implications for phylogenomic data analysis and interpretation.

Main Results:

• Unmodeled genealogical heterogeneity can systematically bias substitution model selection, favoring overly complex models.
• This bias can lead to spurious support for parameter-rich models even with simple underlying evolutionary processes.
• Consequences include distorted phylogenetic estimates and unreliable divergence times.

Conclusions:

• Substitution model choice in phylogenomics must explicitly account for genealogical heterogeneity.
• Genealogical heterogeneity can serve as a diagnostic for violations of core phylogenetic assumptions.
• Integrating gene-tree discordance into model selection is essential for robust phylogenetic inference and biological interpretation.