Performance of standard and stochastic branch-site models for detecting positive selection among coding sequences

Area of Science:

• Evolutionary biology
• Molecular evolution
• Phylogenetics

Background:

• The standard branch-site model analyzes natural selection on coding sequences but requires prior knowledge of lineages under selection.
• This prior information is often unavailable, limiting the standard model's applicability in evolutionary studies.
• A stochastic branch-site model was developed to handle variability in selection patterns without prior lineage information.

Purpose of the Study:

• To compare the performance of the standard and stochastic branch-site models in detecting positive selection.
• To assess the sensitivity and specificity of both models using extensive simulations.
• To evaluate their utility in different scenarios, particularly when prior lineage information is known or unknown.

Main Methods:

• Extensive simulations were conducted to compare the standard and stochastic branch-site models.
• Sensitivity and specificity of tests for positive selection were evaluated under various conditions.
• Comparisons were made against the mixed-effects model of evolution.

Main Results:

• Both models exhibit low Type I error rates, being conservative under neutral or negative selection.
• The standard model is more powerful when prior lineage information is accurate.
• The stochastic model outperforms the standard model when prior lineage information is incorrect or unavailable.

Conclusions:

• The standard branch-site model is suitable for confirmatory analyses requiring known foreground lineages.
• The stochastic branch-site model is preferable for exploratory studies, especially when prior lineage information is uncertain.
• The stochastic approach shows advantages over the standard and mixed-effects models in exploratory contexts.