StarBeast3

StarBeast3 performs Bayesian inference of the multispecies coalescent (MSC) model using Markov chain Monte Carlo (MCMC) to jointly infer species trees and embedded gene trees from multilocus genomic data.

Key Features:

• Bayesian Multispecies Coalescent Framework: Employs a Bayesian MSC framework to jointly infer gene trees embedded within a species tree, addressing bias in branch lengths, substitution rates, and topology arising from concatenation.
• Efficient MCMC Operators: Implements advanced proposal kernels and adaptive MCMC operators to increase sampling efficiency compared to StarBeast2 and *BEAST.
• Relaxed Clock Model Efficiency: Optimized for relaxed molecular clock models that allow variation in evolutionary rates across tree branches.
• Parallelised Gene Tree Inference: Leverages parallel computation for gene tree inference to scale to large datasets, including analyses with 100 or more genes.
• Performance and Scalability: Demonstrates up to ~1.5 orders of magnitude speed improvement over StarBeast2 and over two orders of magnitude improvement over *BEAST, depending on dataset and parameters.

Scientific Applications:

• Species tree estimation: Accurate inference of species trees from multilocus genomic data while accounting for gene tree discordance under the MSC.
• Phylogenomics of large datasets: Analysis of extensive gene sampling across multiple species, including datasets with 100+ loci.
• Evolutionary rate inference: Studies using relaxed clock models to assess among-branch substitution rate variation.
• Complex phylogenetic inference: Estimation of branch lengths, substitution rates, and topology in scenarios where concatenation introduces bias.

Methodology:

Performs Bayesian MCMC sampling of the MSC by embedding gene trees within a species tree and conducting simultaneous inference under the Bayesian framework, using advanced proposal kernels/adaptive MCMC operators and parallelised gene-tree inference, and supporting relaxed molecular clock models.