MCMC

MCMC implements Metropolis-coupled Markov chain Monte Carlo (MCMC) for Bayesian phylogenetics in BEAST 2 via the CoupledMCMC package, enabling more efficient exploration of posterior distributions of complex evolutionary models.

Key Features:

• Parallel Processing: Runs multiple MCMC chains in parallel with heated chains and a cold chain to enhance exploration of the phylogenetic state space.
• Adaptive Temperature Tuning: Dynamically adjusts temperature differences between heated chains to achieve a target acceptance probability for state exchanges among chains.
• BEAST 2 Integration: Integrated into BEAST 2 through the CoupledMCMC package to perform Metropolis-coupled MCMC within BEAST 2's Bayesian phylogenetic framework.

Scientific Applications:

• Phylogenetic Analysis: Facilitates Bayesian inference under complex evolutionary models by improving posterior sampling for phylogenetic relationships.
• Multi-core CPU Utilization: Exploits multi-core CPUs via parallel chains to accelerate computationally intensive Bayesian phylogenetic analyses.

Methodology:

Runs multiple chains at varying temperatures using Metropolis-coupled MCMC, allows heated chains to propose states for other chains including the cold chain, and adaptively tunes temperature differences based on a target acceptance probability for exchanges.