HSMRF

HSMRF models time-varying birth rates in piecewise-constant birth-death phylogenetic frameworks using a horseshoe Markov random field prior to detect and quantify diversification and phylodynamic rate shifts.

Key Features:

• Locally Adaptive Bayesian Framework: Performs joint Bayesian estimation of phylogeny, diversification parameters, and nuisance parameters and is implemented in RevBayes.
• Flexible Time-Varying Models: Uses horseshoe Markov random field (HSMRF) priors within piecewise-constant birth-death models to capture both slow and rapid temporal changes in birth rates, offering greater flexibility relative to Gaussian Markov random field (GMRF) priors.
• High Precision and Accuracy: Demonstrates higher precision than GMRF-based models in comparative studies while maintaining accuracy and distinguishing true signal from stochastic noise in birth-death processes.
• Application to Diverse Datasets: Validated on simulated diversification scenarios and empirical datasets, including detection of rapid speciation-rate decreases in the Australian gecko family Pygopodidae and elevated infection-rate periods in HIV subtype A in Russia and Ukraine.

Scientific Applications:

• Macroevolutionary Analysis: Models speciation as birth events to infer historical diversification dynamics and detect speciation-rate shifts, exemplified by studies on Pygopodidae.
• Phylodynamics of Infectious Diseases: Models new infections as birth events to infer changes in transmission rates over time, exemplified by analyses of HIV subtype A in Russia and Ukraine.

Methodology:

Implements a piecewise-constant birth-death model with an integrated horseshoe Markov random field prior and performs joint Bayesian estimation of phylogeny, diversification parameters, and nuisance parameters in RevBayes; comparative analyses with GMRF-based models assess precision and accuracy, and the piecewise-HSMRF framework balances flexibility and computational efficiency.