Distinguishing positive selection from neutral evolution: boosting the performance of summary statistics

Area of Science:

• Population Genetics
• Evolutionary Biology
• Statistical Genomics

Background:

• Traditional summary statistics in population genetics lack the capacity to fully capture information for distinguishing evolutionary hypotheses.
• Existing neutrality tests may have limitations in detecting specific evolutionary events like selective sweeps or demographic changes.

Purpose of the Study:

• To develop and evaluate a boosting-based statistical method for enhanced detection of selective sweeps in population genetics.
• To assess the performance of this boosting method against established neutrality tests and its robustness against demographic events like bottlenecks.
• To identify the relative contributions of different summary statistics (e.g., integrated haplotype homozygosity, Tajima's π, Watterson's θ) in detecting selection and demographic events.

Main Methods:

• Application of a boosting algorithm, a machine learning technique that combines multiple simple classifiers to maximize predictive accuracy.
• Implementation of boosting for analyzing population genetic data to detect signals of positive selection (selective sweeps).
• Comparative analysis of the boosting method's power against existing neutrality tests and its false positive rate under demographic scenarios such as population bottlenecks.

Main Results:

• The boosting implementation demonstrates high power in detecting selective sweeps.
• Demographic events, specifically bottlenecks, did not lead to a significant excess of false positives, indicating robustness.
• The boosting method performs favorably when compared to other established neutrality tests.
• Integrated haplotype homozygosity was found to be highly informative for recent selective sweeps, while Tajima's π was better for older sweeps.
• Watterson's θ contributed the most information for differentiating between bottlenecks and selection.

Conclusions:

• Boosting offers a powerful and robust statistical framework for detecting selective sweeps in population genetics.
• The method effectively distinguishes between selection and demographic events, providing valuable insights into evolutionary processes.
• Different summary statistics have varying utility depending on the age of selective sweeps, with Watterson's θ being crucial for distinguishing selection from bottlenecks.