Fast and consistent estimation of species trees using supermatrix rooted triples.

Area of Science:

• Phylogenetics and evolutionary biology
• Computational biology and bioinformatics

Background:

• Concatenated sequence alignments are standard for inferring species-level relationships.
• Maximum likelihood (ML) analysis of concatenated data can yield misleading species trees when gene trees conflict due to incomplete lineage sorting.

Purpose of the Study:

• To develop a novel, computationally efficient method for species tree reconstruction.
• To address limitations of standard phylogenetic methods when gene tree topologies differ.

Main Methods:

• Developed the SuperMatrix Rooted Triple (SMRT) method, utilizing a modified mincut supertree algorithm.
• Inferred rooted triples from concatenated alignments, with SMRT-ML using ML for triple inference.
• Evaluated SMRT-ML performance via simulations (Jukes-Cantor, general time-reversible models) and an empirical yeast gene dataset.

Main Results:

• SMRT-ML successfully converged to the correct species tree in scenarios where ML on concatenated data failed.
• The method demonstrated statistical consistency, particularly under the multispecies coalescent model.
• SMRT-ML can be conservative, sometimes yielding partially unresolved trees for problematic clades.

Conclusions:

• SMRT provides a statistically consistent and computationally efficient estimator for species trees.
• The method is robust to gene tree discordance caused by incomplete lineage sorting.
• SMRT offers an improved approach for phylogenetic inference in complex evolutionary scenarios.