Evaluating the Performance of Widely Used Phylogenetic Models for Gene Expression Evolution

Area of Science:

• Evolutionary biology
• Genomics
• Bioinformatics

Background:

• Phylogenetic comparative methods (PCMs) are vital for studying gene expression evolution.
• The applicability of PCMs, developed for quantitative traits, to gene expression data remains unclear.
• Model fit is crucial for reliable conclusions in phylogenetic comparative studies of gene expression.

Purpose of the Study:

• To evaluate the realism of distributional assumptions for phylogenetic models using gene expression data.
• To assess the performance of various phylogenetic models in describing gene expression datasets.
• To identify factors influencing model performance in phylogenetic comparative expression studies.

Main Methods:

• Fitted multiple phylogenetic models of trait evolution to 8 published gene expression datasets (54,774 genes, 145,927 gene-tissue combinations).
• Employed a validated approach to assess the absolute goodness-of-fit for the best-performing model.
• Analyzed model performance using statistical tests to identify deviations from assumptions.

Main Results:

• Ornstein-Uhlenbeck models, reflecting stabilizing selection, were preferred for 66% of gene-tissue combinations.
• The best-fit model demonstrated good performance for 61% of gene-tissue combinations.
• Poor model performance was often linked to uncaptured heterogeneity in evolutionary rates.

Conclusions:

• Phylogenetic models, particularly Ornstein-Uhlenbeck, are frequently suitable for gene expression data.
• A significant portion of gene expression datasets are not adequately described by current best-fit models.
• Assessing model performance is essential for robust phylogenetic comparative analyses of gene expression and for driving new model development.