Estimating Phylogenies from Shape and Similar Multidimensional Data: Why It Is Not Reliable

Area of Science:

• Evolutionary Biology
• Computational Biology
• Phylogenetics

Background:

• Controversy exists regarding the use of multidimensional data (geometric morphometrics, gene expression) for phylogenetic estimation.
• Simulations are crucial for understanding factors influencing the reliability of phylogenetic inference from complex datasets.

Purpose of the Study:

• To investigate the reliability of phylogenetic estimation using multidimensional phenotype data.
• To identify key evolutionary factors impacting phylogenetic accuracy.

Main Methods:

• Simulations of evolution in multidimensional phenotype spaces with varying dimensionality and evolutionary models.
• Comparison of phylogenetic methods including squared-change parsimony, maximum likelihood, Wagner, Euclidean parsimony, neighbor-joining, and UPGMA.
• Analysis of phylogenies with 4 to 8 taxa under Brownian motion and stabilizing selection models.

Main Results:

• Squared-change parsimony and maximum likelihood outperformed other methods.
• High dimensionality under isotropic Brownian motion supports reliable phylogeny estimation.
• Phenotypic integration and stabilizing selection significantly reduce phylogenetic estimate reliability.
• Effects of dimensionality and integration extend to phylogenies with more than four taxa.

Conclusions:

• Multidimensional data can be unreliable for phylogenetic estimation under plausible evolutionary models.
• Phenotypic integration and specific evolutionary pressures (e.g., stabilizing selection) pose significant challenges.
• Careful consideration of evolutionary models and data properties is essential for accurate phylogenetic inference.