Text

theses

Viruses infect a wide variety of hosts across all domains of life. Despite their ubiquity, and a long history of virus research, fundamental questions such as what constitutes a virus species, and how viruses evolve and are modeled, have yet to be adequately answered. We compared viral sequences across five genomic architectures (single- and double-stranded DNA and RNA) and demonstrate the presence of substitution bias, especially in single-stranded viruses. Most striking is a consistent pattern of over-represented cytosine-to- thymine substitutions in single-stranded DNA (ssDNA) viruses. This led us to question the validity of using time-reversible nucleotide substitution models in viral phylogenetic inference, as these models assume equal rates of forward and reverse substitutions. We found that an unrestricted substitution model fit the data better for most single- and double-stranded viral datasets, as measured by corrected Akaike Information Criterion, and hierarchical likelihood ratio test scores. We also approached the question of virus species identification by examining members of the most species-rich viral genus Begomovirus (Family Geminiviridae), which are circular single-stranded DNA (ssDNA) viral crop pathogens transmitted by whitefly vectors. We used novel sweet potato-infecting begomoviruses (sweepoviruses) collected during a recent vector-enabled metagenomic survey to evaluate the concept of pairwise percent nucleotide identity threshold as a criterion for species demarcation. We demonstrate that species demarcation based on pairwise percent nucleotide identities group divergent sweepovirus clusters together, and is highly influenced by when, and how much sampling occurs.

ETD_6812

Ph.D.

Includes bibliographical references

by Yee Mey Seah

doi:10.7282/T31C1ZVC

ETD doctoral

The author owns the copyright to this work.