Species Tree Inference from Gene Splits by Unrooted STAR Methods.

Area of Science:

• Phylogenetics
• Computational Biology
• Evolutionary Biology

Background:

• Inferring species tree topology from gene trees is a key challenge in phylogenetics.
• Previous methods, like Liu and Yu's, showed promise but lacked theoretical proof for larger datasets.
• Statistical consistency under the multispecies coalescent model was limited to four-taxon trees.

Purpose of the Study:

• To mathematically prove the statistical consistency of the Liu and Yu species tree inference method for an arbitrarily large number of taxa.
• To extend the theoretical understanding of phylogenetic inference methods.
• To explore the applicability of the method beyond small, simulated datasets.

Main Methods:

• Developed a novel theoretical approach connecting to a generalization of the STAR method.
• Analyzed the statistical properties of the method under the multispecies coalescent model.
• Focused on the distribution of splits within gene trees, rather than individual gene tree topologies.

Main Results:

• Established the statistical consistency of the Liu and Yu method for arbitrarily large species trees.
• Demonstrated that the method's consistency relies on the distribution of splits, not individual gene tree topologies.
• Provided a theoretical foundation for the method's observed good performance in simulations.

Conclusions:

• The Liu and Yu method is statistically consistent for inferring species trees from gene trees, regardless of the number of taxa.
• The method's reliance on split distributions offers a robust approach to phylogenetic inference.
• Further considerations for handling multiple samples per taxon per gene are discussed for enhanced statistical consistency.