Multinucleotide mutations cause false inferences of lineage-specific positive selection

Area of Science:

• Evolutionary biology
• Genomics
• Molecular evolution

Background:

• Phylogenetic tests like the branch-site test (BST) assume single, independent nucleotide substitutions.
• Multinucleotide mutations (MNMs), affecting adjacent sites, are common during DNA replication and often non-synonymous.
• MNMs can potentially confound evolutionary analyses by mimicking signals of positive selection.

Purpose of the Study:

• To investigate whether the BST misinterprets sequence patterns generated by MNMs as evidence of positive selection.
• To evaluate the impact of realistic MNM rates on BST inferences of adaptive evolution.
• To assess the reliability of published BST results in light of potential MNM-induced biases.

Main Methods:

• Analysis of two genome-scale datasets from mammals and flies.
• Evaluation of codons with multiple differences in BST analyses.
• Simulations using empirically derived conditions to model MNM effects without positive selection.

Main Results:

• Codons with multiple differences accounted for nearly all BST support for lineage-specific positive selection.
• Simulations demonstrated that realistic MNM rates introduce a strong bias towards false positive selection inferences.
• Empirically derived conditions showed MNMs can cause false positives at rates comparable to actual BST findings.

Conclusions:

• The BST cannot reliably distinguish between adaptive evolution and neutral fixation of MNMs.
• Many published inferences of adaptive evolution using the BST may be artefacts of model violation due to unaddressed MNMs.
• A new model incorporating MNMs is proposed to mitigate this bias in evolutionary analyses.