Evaluating the Performance of Probabilistic Algorithms for Phylogenetic Analysis of Big Morphological Datasets: A...

Area of Science:

• Evolutionary biology
• Phylogenetics
• Computational biology

Background:

• Reconstructing the tree of life requires accurate phylogenetic inference for extant and extinct organisms, with morphological data crucial for extinct species.
• Probabilistic methods like maximum likelihood and Bayesian inference are increasingly used in morphological phylogenetics, replacing traditional parsimony methods.
• A lack of benchmark assessments for probabilistic software on morphological data, especially with large datasets, hinders method selection.

Purpose of the Study:

• To evaluate the performance of major probabilistic phylogenetic software using morphological data.
• To compare Bayesian inference and maximum likelihood methods under varying taxonomic sampling and missing data conditions.
• To provide guidance on software selection for morphological phylogenetic analyses.

Main Methods:

• Tested four probabilistic software packages: MrBayes (Bayesian), RevBayes (Bayesian), IQ-TREE (Maximum Likelihood), and RAxML (Maximum Likelihood).
• Analyzed performance using simulated morphological datasets with varied taxonomic sampling and missing data percentages.
• Evaluated tree reconstruction accuracy using Robinson-Foulds (RF), Matching Splits (MS), and Kuhner-Felsenstein (KF) distances.

Main Results:

• Increased taxonomic sampling consistently improved accuracy, precision, and resolution across all tested software.
• Bayesian inference programs (MrBayes, RevBayes) showed superior consistency, accuracy, and robustness, particularly with high missing data, under the RF metric.
• The MS metric favored IQ-TREE's more resolved topologies, while RF metric penalized false positives more heavily than MS metric penalized polytomies.

Conclusions:

• Bayesian inference is recommended over maximum likelihood for morphological data analysis to avoid false positives.
• Increased taxonomic sampling is crucial for improving phylogenetic reconstruction accuracy regardless of the software used.
• The choice of tree comparison metric (e.g., RF vs. MS) influences the perceived performance of different phylogenetic methods.