Parallel Metropolis Coupled Markov Chain Monte Carlo for Isolation with Migration Model

Isolation with Migration model (IM), which jointly estimates divergence times and migration rates between two populations from DNA sequence data, can capture many phenomena when one population splits into two. The parameters inferences for IM are based on Markov Chain Monte Carlo method (MCMC). Standard implementations of MCMC are prone to fall into local optima. Metropolis Coupled MCMC [(MC)(3)] as a variant of MCMC can more readily explore multiple peaks in posterior distribution of trees. Expensive execution time has limited the application of (MC)(3). This paper proposes a Parallel Metropolis Coupled Markov Chain Monte Carlo for IM. The proposed parallel algorithm retains the ability of (MC)(3) and maintains a fast execution time. Performance results indicate nearly linear speed improvement. This paper provides researcher with rapider and more high-efficiency methods to study population genetics and molecular ecology problems aided with super computer.