Fundamental differences between the methods of maximum likelihood and maximum posterior probability in phylogenetics.

Area of Science:

• Evolutionary Biology
• Computational Biology
• Phylogenetics

Background:

• Phylogenetic tree reconstruction is crucial for understanding evolutionary relationships.
• Maximum likelihood and Bayesian inference are common methods for inferring evolutionary trees.
• Previous work suggested potential equivalence between bootstrap frequencies and posterior probabilities under specific conditions.

Purpose of the Study:

• To investigate discrepancies between maximum likelihood and Bayesian inference methods in phylogenetic tree topology reconstruction.
• To analyze the impact of differing inference methods on tree support metrics.
• To clarify the conditions under which bootstrap frequencies approximate posterior probabilities.

Main Methods:

• Utilized a four-taxon example with a simple evolutionary model.
• Employed maximum likelihood and Bayesian inference (maximum posterior probability) methods.
• Examined the informativeness of data patterns under each method.

Main Results:

• Maximum likelihood and Bayesian inference can produce different optimal tree topologies.
• Data patterns uninformative in maximum likelihood can be informative in Bayesian inference.
• Bootstrap frequencies and posterior probabilities are not always approximately equal, contrary to some prior interpretations.

Conclusions:

• The choice of phylogenetic inference method (maximum likelihood vs. Bayesian) significantly affects tree topology and support values.
• The assumption of approximate equality between bootstrap frequencies and posterior probabilities requires careful consideration of prior distributions.
• Accurate phylogenetic inference necessitates understanding the distinct properties of different statistical methods.