Author's School

Graduate School of Arts & Sciences

Author's Department/Program

Biology and Biomedical Sciences: Computational and Systems Biology

Language

English (en)

Date of Award

5-24-2012

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Chair and Committee

David Wang

Abstract

Pathogen detection is a challenging problem given that any given specimen may contain one or more of many different microbes. Additionally, a specimen may contain microbes that have yet to be discovered. Traditional diagnostics are ill-equipped to address these challenges because they are focused on the detection of a single agent or panel of agents. I have developed three innovative computational approaches for analyzing high-throughput genomic assays capable of detecting many microbes in a parallel and unbiased fashion. The first is a metagenomic sequence analysis pipeline that was initially applied to 12 pediatric diarrhea specimens in order to give the first ever look at the diarrhea virome. Metagenomic sequencing and subsequent analysis revealed a spectrum of viruses in these specimens including known and highly divergent viruses. This metagenomic survey serves as a basis for future investigations about the possible role of these viruses in disease. The second tool I developed is a novel algorithm for diagnostic microarray analysis called VIPR: Viral Identification with a PRobabilistic algorithm). The main advantage of VIPR relative to other published methods for diagnostic microarray analysis is that it relies on a training set of empirical hybridizations of known viruses to guide future predictions. VIPR uses a Bayesian statistical framework in order to accomplish this. A set of hemorrhagic fever viruses and their relatives were hybridized to a total of 110 microarrays in order to test the performance of VIPR. VIPR achieved an accuracy of 94% and outperformed existing approaches for this dataset. The third tool I developed for pathogen detection is called VIPR HMM. VIPR HMM expands upon VIPR's previous implementation by incorporating a hidden Markov model: HMM) in order to detect recombinant viruses. VIPR HMM correctly identified 95% of inter-species breakpoints for a set of recombinant alphaviruses and flaviviruses Mass sequencing and diagnostic microarrays require robust computational tools in order to make predictions regarding the presence of microbes in specimens of interest. High-throughput diagnostic assays coupled with powerful analysis tools have the potential to increase the efficacy with which we detect pathogens and treat disease as these technologies play more prominent roles in clinical laboratories.

Comments

Permanent URL: http://dx.doi.org/10.7936/K73776Q4

Download

Share

COinS