PBPI: a high performance implementation of Bayesian phylogenetic inference

摘要: This paper describes the implementation and performance of PBPI, a parallel Bayesian phylogenetic inference method for DNA sequence data. By combining Markov Chain Monte Carlo (MCMC) with likelihood-based assessment phylogenies, inferences can incorporate complex statistic models into process tree estimation. However, analyses are extremely computationally expensive. PBPI uses algorithmic improvements processing to achieve significant improvement over comparable programs. We evaluated accuracy using simulated dataset on System X, terascale supercomputer at Virginia Tech. Our results show that identifies equivalent estimates 1424 times faster 256 processors than widely-used, best-available (albeit sequential), program. also achieves linear speedup number large problem sizes. Most importantly, framework enables analysis datasets previously impracticable.