Date of Award
Doctor of Philosophy
Karin S. Dorman
There is a rich tradition in mathematical biology of modeling virus population dynamics within hosts. Such models can reproduce trends in the progression of viral infections such as HIV-1, and have also generated insights on the emergence of drug resistance and treatment strategies. Existing mathematical work has focused on the problem of predicting dynamics given model parameters. The problem of estimating model parameters from observed data has received little attention. One reason is likely the historical difficulty of obtaining high-resolution samples of virus diversity within hosts. Now, next-generation sequencing (NGS) approaches developed in the past decade can supply such data.
This thesis presents two Bayesian methods that harness classical models to generate testable hypotheses from NGS datasets. The quasispecies equilibrium explains genetic variation in virus populations as a balance between mutation and selection. We use this model to infer fitness effects of individual mutations and pairs of interacting mutations. Although our method provides a high resolution and accurate picture of the fitness landscape when equilibrium holds, we demonstrate the common observation of populations with coexisting, divergent viruses is unlikely to be consistent with equilibrium. Our second statistical method estimates virus growth rates and binding affinity between viruses and antibodies using the generalized Lotka-Volterra model. Immune responses can explain coexistence of abundant virus variants and their trajectories through time. Additionally, we can draw inferences about immune escape and antibody genetic variants responsible for improved virus recognition.
Emily Anne King
King, Emily Anne, "Bayesian inference of virus evolutionary models from next-generation sequencing data" (2017). Graduate Theses and Dissertations. 15337.