Multiple hypothesis testing to detect lineages under positive selection that affects only a few sites.

Area of Science:

• Evolutionary biology
• Genomics
• Bioinformatics

Background:

• Detecting positive Darwinian selection is crucial with increasing genomic data.
• Branch-site models identify short episodes of molecular adaptation but require specifying branches a priori.
• Testing multiple branches necessitates correction for multiple testing.

Purpose of the Study:

• Evaluate the performance of six multiple test correction procedures for branch-site models.
• Assess the reliability and power of these corrections when testing every phylogenetic branch for positive selection.

Main Methods:

• Computer simulations were used to evaluate six multiple test correction procedures.
• Four methods controlled familywise error rates (FWER), and two controlled false discovery rates (FDR).
• Performance was assessed under various conditions, including sequence divergence and model violations, using gene sequences from the CD2 gene.

Main Results:

• All tested correction procedures maintained acceptable FWER, except with highly divergent sequences or severe model violations.
• Test power to detect positive selection varied with selection strength and sequence divergence, peaking at intermediate levels.
• FWER-controlling procedures showed similar power; FDR-controlling methods offered slightly more power but higher FWER.

Conclusions:

• Multiple test correction procedures are valuable for branch-site model analyses when numerous branches require testing.
• Rom's procedure and the Bonferroni correction are recommended for controlling FWER.
• FDR-controlling methods provide a power advantage with a slight increase in FWER, suitable for exploratory analyses.