Comparative evaluation of maximum parsimony and Bayesian phylogenetic reconstruction using empirical morphological...

Area of Science:

• Evolutionary Biology
• Phylogenetics
• Computational Biology

Background:

• Discrete morphological data are increasingly used in Bayesian phylogenetics (BI), particularly in total evidence analysis and Bayesian molecular dating.
• Current BI methods assume probabilistic Markov models adequately capture morphological evolution complexity.
• Previous performance evaluations of these models relied heavily on simulated data.

Purpose of the Study:

• To evaluate Bayesian phylogenetic inference (BI) using empirical morphological data under the Lewis' Mk model.
• To compare BI performance against the maximum parsimony (MP) algorithm using real-world datasets.
• To identify factors influencing topological differences between BI and MP phylogenetic trees.

Main Methods:

• Surveyed MorphoBank to collect a large number of empirical morphological matrices.
• Applied Bayesian phylogenetic inference (BI) under the Lewis' Mk model.
• Compared BI results with those obtained from the maximum parsimony (MP) algorithm.

Main Results:

• Trees inferred by BI and MP frequently differed topologically.
• BI generated a higher proportion of polytomic tree topologies compared to MP.
• The number of terminals in morphological matrices was the primary factor associated with topological distance between BI and MP trees; sample size did not significantly influence differences.

Conclusions:

• Bayesian phylogenetic inference (BI) using discrete morphological data is less precise than maximum parsimony (MP).
• The complexity of morphological evolution and model assumptions in BI may lead to less robust phylogenetic reconstructions with empirical data.
• Findings support recent research suggesting MP may be more reliable for inferring phylogenies from discrete morphological datasets.