Homoplasy-Based Partitioning Outperforms Alternatives in Bayesian Analysis of Discrete Morphological Data

Area of Science:

• Evolutionary biology
• Phylogenetics
• Computational biology

Background:

• Bayesian phylogenetic inference using morphological data is gaining traction, especially for time-calibrated analyses with fossils.
• Modeling among-character rate variation (ACRV) is crucial for improving phylogenetic accuracy, similar to molecular data analyses.
• Current methods for handling ACRV, like hidden Markov models or data partitioning, have limitations and lack consensus.

Purpose of the Study:

• To evaluate novel strategies for partitioning morphological data to effectively model among-character rate variation (ACRV).
• To compare the performance of homoplasy-based partitioning against other methods like anatomical partitioning and PartitionFinder2.
• To assess the utility of the stepping-stone method for marginal likelihood approximation in model selection for morphological phylogenetics.

Main Methods:

• Applied Bayesian phylogenetic inference to three discrete morphological datasets.
• Utilized the stepping-stone method for marginal likelihood approximation to compare different partitioning strategies.
• Sorted characters by homoplasy scores derived from implied weighting parsimony analysis for data partitioning.

Main Results:

• Partitioning morphological data by homoplasy scores significantly outperformed other tested partitioning strategies (anatomically-based and PartitionFinder2).
• Homoplasy-based partitioning effectively segregated characters by evolutionary rates, obviating the need for within-partition ACRV modeling.
• Rate multipliers adequately accommodated among-partition rate variation, simplifying the phylogenetic model.

Conclusions:

• Partitioning morphological data by homoplasy is a powerful and easily implementable strategy for addressing ACRV in complex datasets.
• This approach enhances phylogenetic inference accuracy and can be applied to both morphological and potentially molecular data.
• The findings offer practical guidelines for optimizing phylogenetic analyses of discrete morphological data.