Phylogenomic estimation - the science of calculating evolutionary trees from genomic data - is an important biological problem. As the amount of genomic data in biological datasets increases, new methods are needed to analyze this data. Cutting edge analyses may utilize genomes from tens of thousands of species.I present several methods for supertree and species tree estimation: ASTRID, FastRFS, SVDquest, and SIESTA. ASTRID can be used for both species tree and supertree estimation, and is designed to scale to very large datasets while maintaining a high level of accuracy. FastRFS is a supertree method that uses an exact constrained optimization algorithm to find accurate supertrees. SVDquest is a coalescent-aware species tree estimation method that estimates trees directly from sequences without using gene trees. Finally, SIESTA is a modification to the algorithms used by FastRFS, SVDquest, and other methods including ASTRAL that allows for the output and analysis of multiple optimal solutions, if they exist.For all these methods, I describe the algorithms used, along with a theoretical analysis of their running time and their statistical consistency. I also show results on biological and simulated data that demonstrate these methods’ effectiveness over a wide range of model conditions. In addition, I present the results of an experiment that compares various methods on trees simulated under both incomplete lineage sorting (ILS) as well as horizontal gene transfer (HGT).
PDF
download
878987797
028-85220240
