Recoding Amino Acids to a Reduced Alphabet may Increase or Decrease Phylogenetic Accuracy

Area of Science:

• Phylogenetics
• Molecular Evolution
• Bioinformatics

Background:

• Amino acid data presents challenges in phylogenetic reconstruction due to long branches and compositional heterogeneity.
• Recoding alignments to reduced alphabets is a common strategy to mitigate these issues.

Purpose of the Study:

• To evaluate the effectiveness of different amino acid recoding strategies on phylogenetic topological accuracy using simulated data.
• To compare recoding methods based on amino acid exchangeability versus those reducing compositional heterogeneity.

Main Methods:

• Simulated four-taxon tree alignments with varying degrees of branch length and compositional heterogeneity.
• Tested three exchangeability-based recoding methods and one Chi-squared statistic-based recoding method.
• Analyzed accuracy using homogeneous phylogenetic models and compared with advanced models (NDCH, CAT).

Main Results:

• Exchangeability-based recoding generally improved accuracy on compositionally heterogeneous alignments.
• Chi-squared recoding showed variable results, sometimes decreasing accuracy.
• Advanced models (NDCH, CAT) analyzing unrecoded, heterogeneous data often outperformed recoded data with homogeneous models.

Conclusions:

• Amino acid recoding can be beneficial but requires cautious interpretation.
• The effectiveness of recoding depends on the type of heterogeneity and the recoding scheme.
• Developing and utilizing better-fitting models (e.g., NDCH, CAT) is a more robust approach for analyzing complex molecular data.