Large datasets and Trichoptera phylogenetics: DNA barcodes, partitioned phylogenetic models, and the evolution of Phryganeidae.

摘要

Large datasets in phylogenetics---those with a large number of taxa, e.g. DNA barcode data sets, and those with a large amount of sequence data per taxon, e.g. data sets generated from high throughput sequencing---pose both exciting possibilities and interesting analytical problems. The analysis of both types of large datasets is explored in this dissertation. First, the use of DNA barcodes in phylogenetics is investigated via the generation of phylogenetic trees for known monophyletic clades. Barcodes are found to be useful in shallow scale phylogenetic analyses when given a well-supported scaffold on which to place them. One of the analytical challenges posed by large phylogenetic datasets is the selection of appropriate partitioned models of molecular evolution. The most commonly used model partitioning strategies can fail to characterize the true variation of the evolutionary process and this effect can be exacerbated when applied to large datasets. A new, scalable algorithm for the automatic selection of partitioned models of molecular evolution is proposed with an eye toward reducing systematic error in phylogenomics. The new algorithm is tested on a range of empirical datasets and found to provide a better fit of the model to the data as measured by information theoretic metrics like AICc. Indeed, the algorithm is found to perform particularly well when applied to a phylogenomic dataset consisting of ultra-conserved elements (UCEs). Finally, the phylogeny of Phryganeidae is estimated using a large dataset generated using targeted enrichment and high throughput sequencing. Trees generated from different modeling strategies give incongruent, but strongly supported results. The differences between the trees are examined and a new hypothesis for the relationships among the genera within Phryganeidae is posited.