Using MapReduce to Compare Large Collections of Phylogenetic Trees

Abstract

Phylogenetic trees depict the relationships between sets of organisms sharing a common ancestor. In several recent court cases, phylogenetic trees have been critical in convicting and exonerating suspects of purposefully transmitting the HIV virus. Parallel computing applications such as MapReduce Speeds Up Robinson-Foulds (MrsRF) make it possible to compare large collections of phylogenetic trees efficiently, and can be used to ensure that the searches that produce these trees properly converged. In this paper, we describe MrsRF++, an updated version of MrsRF that utilizes the Phoenix++ MapReduce engine. The original version of MrsRF was implemented with Phoenix 1.0, which had a number of critical limitations. Implementing MrsRF with Phoenix++ overcomes many of these shortcomings, and increased program modularity. We ran MrsRF++ on the same two large biological tree sets described in the original MrsRF paper. Our preliminary results indicate that using the Phoenix++ MapReduce engine increases MrsRF's scalability, a fact that will enable scientists to analyze increasingly large collections of phylogenetic trees.