Evaluating Parallel Approaches for Evolutionary Trees
One of Charles Darwin’s main visions was reconstructing a “Tree of Life” showing evolutionary relationships among all species on the planet. Recently, evolutionary biologists have been using new genomic data to reconstruct many parts of the tree of life. However, determining relationships among closely related species remains one of the greatest challenges, especially because closely related species have evolved very few species-specific differences in their DNA sequences. Our research group is using multiple gene regions from different parts of the genome to reconstruct these evolutionary family trees – phylogenies. Many new “species tree” algorithms have been developed to analyze such data for recently diverged species. We work on a genus of birds that has thirty species, plus multiple subspecies. Some of the new programs that we are using (e.g., BEST) require weeks to run on computer clusters for just five or ten species. To scale up our analyses to fifty or more species and subspecies, we need much more computationally efficient approaches, including parallelization. A newer program (*BEAST), which may allow parallel analyses, has just been released.