Improving Information Propagation in Phylogenetic WorkflowsGuang, AugustcreatorLawrence, CharlesAdvisorDunn, CaseyReaderLewis, PaulReaderBrown University. Department of Applied Mathematicssponsor2018xiv, 100 p.born digitalThesis (Ph. D.)--Brown University, 2018thesesDespite the enormous amount of biological variation and technical uncertainty in sequence data, most phylogenetic workflows propagate a single point estimate throughout the numerous analysis components, and only in a forward direction. This approach relies on three implicit assumptions: (i) the order of the analysis steps is biologically justified, (ii) a Markovian dependency structure exists between analysis components, and (iii) there is low relative entropy between results at each analysis step. There is evidence that these assumptions, in particular low relative entropy, are frequently violated in empirical studies with potential detrimental effects in phylogenetic analyses. In this thesis, I lay out a probabilistic framework that provides a unified perspective to provide context for evaluating priorities for future developments of methods and tools. I then develop a generative model of the natural and technical processes that produce observed genomic reads within the framework that can be used to assess and validate approaches that relax the implicit assumptions. Finally, I explore two ways to accommodate and propagate more information in a phylogenetic workflow. The first way, an HMM profile-sampling approach to genome assembly, relaxes the assumption of low relative entropy in results from the genome assembly analysis component. This approach finds relevant applications to HIV transmission networks. The second way, an iterative approach to identifying and resolving transcriptome assembly errors, capitalizes on the assumption of Markovian dependence.HIV/AIDSProbabilityComputational biologyPhylogenyEnglishRPB2018061510.26300/m4j5-dd88In CopyrightCollection is open for research.dissertations