High-Performance Computing for Statistical Phylogenetics

Statistical phylogenetic analyses based on maximum likelihood (ML) and Bayesian inference are computationally challenging because of the intensive nature of the calculations required. At the core of statistical phylogenetics is the calculation of the likelihood (probability) of the observed molecular sequence character states under a specific model of evolution using a recursive algorithm, and the computation of this calculation comprises most of the running time for analyses. Decreasing the time (wall clock) for this computation is the raison d’être for the BEAGLE library, a parallel computing platform for high-performance calculation of phylogenetic likelihoods that makes efficient use of the fine-scale parallelization capabilities of computer processors, especially graphics processing units (GPUs).

Daniel L. Ayres
Assistant Research Scientist
Center for Bioinformatics and Computational Biology