cubeVB

cubeVB applies Bayesian statistics and Markov Chain Monte Carlo (MCMC) simulations to represent phylogenies as distance matrices within a constrained "cube" subspace for modeling distributions over rooted time trees.

Key Features:

• Hierarchical clustering representation: Represents phylogenetic trees through hierarchical clustering of distance matrices.
• Distance-matrix modeling: Models phylogenies as distance matrices with a multivariate normal distribution over selected matrix entries.
• Cube subspace restriction: Operates within a constrained "cube" subspace that provides a bijective representation of tree distributions while excluding some trees outside the cube.
• Bayesian inference with MCMC: Leverages Bayesian statistics and MCMC simulations for sampling over tree distributions.
• Parameter recovery: Demonstrates recovery of key parameters such as tree height and tree length in simulation studies.
• Improved summary trees: Produces summary trees from the matrix representation intended to improve on maximum clade credibility trees.
• Scalability: Enables scaling of Bayesian phylogenetic inference for larger datasets within the cube representation.

Scientific Applications:

• Evolutionary history reconstruction: Modeling distributions over rooted time trees to infer evolutionary relationships and temporal structure.
• Biodiversity and comparative studies: Assessing genetic relationships and evolutionary dynamics across species using large-scale phylogenetic data.

Methodology:

Implements Bayesian inference via Markov Chain Monte Carlo within the BEAST 2 framework, representing phylogenies as distance matrices constrained to a "cube" subspace and modeling selected matrix entries with a multivariate normal distribution informed by hierarchical clustering.