Erasing errors due to alignment ambiguity when estimating positive selection.

Area of Science:

• Evolutionary biology
• Bioinformatics
• Computational biology

Background:

• Estimating positive selection requires accurate multiple sequence alignments.
• Common alignment software can produce errors leading to high false-positive rates in selection detection.

Purpose of the Study:

• To develop a novel statistical method to eliminate false positives in positive selection detection caused by alignment errors.
• To jointly estimate the degree of positive selection and the sequence alignment under a unified evolutionary model.

Main Methods:

• Developed a Bayesian approach using Markov chain Monte Carlo (MCMC) to integrate over all possible alignments.
• Modeled substitutions and insertions/deletions as sequence changes on a phylogenetic tree with site heterogeneity.
• Introduced a Bayesian branch-site test using Bayes factors and Rao-Blackwellization for accurate estimation.

Main Results:

• The novel method effectively eliminates excess false positives stemming from alignment errors.
• The joint estimation approach maintains high power for detecting positive selection, comparable to using true alignments.
• Software implementing this method (BAli-Phy) demonstrated substantially lower alignment error than other leading alignment tools.

Conclusions:

• Jointly estimating sequence alignment and positive selection is a robust solution to the problem of alignment-induced false positives.
• This integrated approach improves the reliability of detecting evolutionary selection pressures.
• The BAli-Phy software offers a significant advancement in accurate sequence alignment and evolutionary analysis.