J. Parallel Distrib. Comput.

Exponential growth of the data available for molecular sequence analysis causes eminent need for methods to analyze large datasets in reasonable time. In molecular phylogenetics maximum-likelihood methods became very popular despite their vast need for computational power. During the last decades parallel computing has proven to be a valuable way to decrease running time of computationally intensive analyses. In this paper we suggest to parallelize the estimation of parameters for evolutionary models and the quartet puzzling algorithm to reconstruct phylogenetic trees from DNA and protein sequences applying the maximum-likelihood principle. Furthermore, we discuss effects of the different parallel granularities of the algorithms