A basic limitation on inferring phylogenies by pairwise sequence comparisons.

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Distance-based phylogenetic methods, like Neighbor-Joining, are widely used for tree construction.
• These methods rely on pairwise sequence distances, assuming additivity under substitution models.
• Rate variation across sites is often modeled using a gamma distribution.

Purpose of the Study:

• To investigate the identifiability of phylogenetic trees when the gamma distribution shape parameter is unknown.
• To determine if different tree topologies can yield indistinguishable pairwise site pattern distributions.

Main Methods:

• Analysis of pairwise site pattern distributions under various phylogenetic tree topologies.
• Exploration of the impact of unknown gamma distribution shape parameters on tree reconstruction.

Main Results:

• Demonstrated that topologically distinct trees with different shape parameters can produce identical pairwise site pattern distributions.
• Showed this ambiguity arises for any distinct choice of shape parameters, not just specific ones.
• Identified conditions, such as clocklike branch lengths or maximum likelihood methods, that restore identifiability.

Conclusions:

• The accuracy of distance-based phylogenetic methods is compromised when the gamma distribution shape parameter is unknown.
• Ambiguity in tree reconstruction is a significant challenge for these methods.
• Identifiability can be achieved under specific constraints or by employing alternative methods like maximum likelihood.