The COVID-19 pandemic created challenges across biological scales and scientific disciplines, spanning from molecular virology to population-scale epidemiology. This thesis presents a body of work that develops and applies quantitative and computational methods to study SARS-CoV-2 across biological scales, from viral replication within individual hosts to infection control strategies at the population level. Each component of this research arose from a distinct phase of the pandemic and addresses a key scientific challenge of that period.
In the early phase, when no vaccines or antivirals were available, we developed stochastic models of inter-host SARS-CoV-2 infection dynamics to evaluate non-pharmaceutical interventions (NPIs) such as testing, quarantine, and isolation. These models describe infection progression in pre-detectable, infectious, and post-infectious stages and integrate empirical data on viral load dynamics and test performance to estimate transmission risk under different intervention schemes. The models were implemented in the open-source software COVIDStrategyCalculator, enabling policymakers and health authorities to compare NPI strategies in real time. Post-hoc validation with the first SARS-CoV-2 human challenge study confirmed the model’s predictive accuracy, and the framework has since been adapted to other pathogens, including the Monkeypox virus.
With the advent of vaccines, attention shifted from epidemiological control to the characterization of molecular mechanisms of viral replication and immune modulation. The success of mRNA vaccines underscored the importance of RNA chemistry, particularly the stabilizing role of modified nucleosides, in achieving effective expression without triggering excessive innate immunity. This highlighted the broader need to characterize RNA modifications and other epitranscriptomic features that influence viral behavior and vaccine design. Nanopore direct RNA sequencing (dRNA-seq) emerged as a promising method for studying these features, as it allows native RNA molecules to be read directly, including their post-transcriptional modifications. However, the technology posed considerable computational challenges related to signal interpretation and throughput.
This thesis advances nanopore dRNA-seq as a tool for analyzing viral and host RNA biology. A systematic characterization of dRNA-seq signal and sequencing errors defined the limits of current sequencing accuracy and informed new algorithms for detecting RNA modifications. Building on this foundation, a novel barcoding and barcode-inference framework, WarpDemuX, was developed to enable the pooled sequencing of multiple RNA samples within a single run. This method eliminates batch effects, reduces cost and turnaround time, and supports longitudinal and comparative studies of viral transcriptomes. Application of this approach to SARS-CoV-2 revealed time-resolved viral replication patterns, stable subgenomic RNA hierarchies, and dynamic regulation of poly(A) tail lengths during infection. As a methodological stress test, the same framework was extended to the pooled sequencing of tRNAs, demonstrating its generalizability to short and heavily modified RNA molecules.
Taken together, these advances demonstrate how computational modeling and sequencing advances can jointly resolve questions related to viral dynamics and support epidemic response. Mechanistic models convert limited early data into actionable guidance, while sequencing frameworks reveal the RNA-level processes that govern replication and adaptation. In combination, they establish a robust, generalizable foundation for rapid analysis, modeling, and control of emerging RNA viruses, delivering methods that can be applied beyond SARS-CoV-2 and support future global health preparedness and mRNA-based biotechnology.
