694
Parallel Computing to Speed up Whole-Genome Analyses Using Independent Metropolis-Hastings Sampling
Bayesian multiple regression methods are widely used in whole-genome analyses by constructing a Markov chain with a stationary distribution equal to the posterior distribution of unknown parameters. In whole-genome analyses, chains of about 50,000 samples are typically used, for which the computation is intensive. Thus, it is desirable if parallel computing, taking advantage of multiple cores on computers, could be used to speed up Bayesian methods. In this paper, a strategy using Independent Metropolis-Hastings (IMH) sampling to parallelize Markov chain Monte Carlo (MCMC) sampling for whole-genome analyses has been shown. We also propose a strategy to construct the proposal distribution in IMH. Addressing the heavy computational burden associated with Bayesian methods by parallel computing will lead to greater use of these methods. .
Keywords:
Whole-genome analyses
Parallel computing
Independent Metropolis-Hastings Sampling