Scalable and Robust Bayesian Methods for Joint Inference of Pathogen Evolution and Transmission in Outbreaks Restricted; Files Only

Abstract

Phylodynamic methods integrate pathogen genetic sequences with traditional epidemiological data to jointly infer transmission dynamics and evolutionary relationships during infectious disease outbreaks. By capturing patterns of pathogen diversity across hosts, genetic data provide critical resolution where case-based epidemiological data alone fall short. This thesis develops scalable and robust Bayesian phylodynamic frameworks that enhance outbreak reconstruction for large-scale epidemics and enable joint inference in complex settings with multiple co-circulating pathogens. 

In our first topic, we adapt and extend a previous, highly accurate mechanistic and Bayesian phylodynamic model developed by Lau and colleagues, which utilizes a realistic likelihood to systematically integrate epidemiological and genomic data. Our model is designed to overcome the scalability issues of the previous method in large outbreaks. By adopting an infinite-sites model for modeling genetic evolution, we substantially reduce the dimension of the model and achieve a significant improvement in computational efficiency for inference via data-augmentation Markov Chain Monte Carlo (MCMC). We validate our method through simulations and analysis of an outbreak of Foot and Mouth Disease Virus, where it provides results comparable to the original model but at a fraction of the computational cost. 

In our second topic, we focus on outbreaks in which two subtypes of a pathogen circulate within a population, a phenomenon found in many outbreak settings, such as swine influenza in livestock or healthcare-associated outbreaks. We develop a fully Bayesian mechanistic model to analyze the transmission and evolution of two circulating subtypes to jointly infer the dynamics of between-host transmission and within-host evolution. In these outbreak settings, the transmission tree is complicated by multiple types of transmission events, and distinguishing these transmission events is critical for accurate outbreak reconstruction. Inference is performed via a custom data-augmentation MCMC implementation to impute high-dimensional quantities of missing data. We demonstrate the performance of our method via simulation studies and reconstruct ambiguous transmission events with high accuracy. 

In our third topic, we generalize a hierarchical Bayesian model of within-host viral load for two circulating subtypes in an outbreak of H1N1 and H3N2 swine influenza. Understanding the dynamics of viral load during infection may answer open questions around relative transmissibility and duration of infectivity between infection subtypes. Due to the limited observation period for our swine influenza data, for many individuals we may not observe the full viral trajectory. Via reversible-jump MCMC, we implement a model which allows us to infer viral proliferation and clearance, then apply it to a dataset of co-circulating H1N1 and H3N2 swine influenza at a county fair.  

Table of Contents

1 Introduction

1.1 Overview

1.1.1 Phylodynamic Methods

1.1.2 Data-Augmentation Markov Chain Monte Carlo

1.1.3 Systematic Bayesian Integration of Epidemiological and Genetic Data

1.2 Outline of Research

2 Scalable Bayesian inference of transmission tree from epidemiological and genomic data

2.1 Introduction

2.2 Model and Methods

2.2.1 Stochastic Epidemiological Process

2.2.2 Stochastic Evolutionary Process

2.2.3 A Bayesian Modeling Framework

2.3 Results

2.3.1 Simulation Studies

2.3.2 Improved Computational Scalability

2.3.3 Tolerance to Incomplete Genetic Sampling

2.3.4 Impact of Unobserved Infections

2.3.5 Case Study: Foot-and-Mouth Disease Virus (FMDV) Outbreak

in the UK

2.4 Discussion

2.5 Appendix

2.5.1 Supporting Text

2.5.2 Supporting Figures

2.5.3 Supporting Tables

3 Phylodynamic inference for transmission networks and outbreak dy-

namics of co-circulating pathogens

3.1 Introduction

3.2 Motivation

3.3 Methods

3.3.1 Stochastic Epidemic Model

3.3.2 Evolutionary Model

3.3.3 Data-augmentation MCMC Inference

3.4 Results

3.4.1 Simulation Studies

3.4.2 Improvement in inference beyond independent single-infection

inference

3.5 Discussion

3.6 Appendix

3.6.1 Supporting Text

3.6.2 Supporting Figures

3.6.3 Supporting Tables

4 A generalized Bayesian hierarchical approach for modeling temporal

viral load dynamics

4.1 Introduction

4.2 Methods

4.2.1 Viral Trajectory Model

4.2.2 Reversible-Jump MCMC

4.3 Results

4.3.1 Validation of Method Using 2020 SARS-CoV-2 testing data

from the 2019-2020 NBA season

4.3.2 Swine Influenza Outbreak and Dual Infections

4.3.3 Viral Trajectory Parameters

4.3.4 Discussion

4.4 Appendix

4.4.1 Model Likelihood

4.4.2 Reversible-Jump MCMC Implementation

4.4.3 Supporting Figures

4.4.4 Supporting Tables

5 Future Work

5.1 Missing Data and Population-Level Inference

5.2 Modeling and Inference Across Scales

5.3 Comparing Viral Dynamics of Superinfected and Coinfected Swine

Bibliography

About this Dissertation

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Biostatistics
Degree	Ph.D.
Submission	Dissertation
Language	English
Research Field	Biology, Biostatistics Health Sciences, Epidemiology
Keyword	Biostatistics Phylodynamics Infectious disease Epidemiology Reversible Jump MCMC Bayesian inference Data augmentation MCMC
Committee Chair / Thesis Advisor	Lance Waller, Emory University Max Lau, Emory University
Committee Members	Katia Koelle, Emory University Zhaohui Qin, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 21 August 2027	2025-07-10 19:03:35 -0400	File download under embargo until 21 August 2027

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions