PhD Thesis Colloquium: Shreya Chowdhury

 Mathematical and data-driven modelling of viral infections: Applications to HIV/AIDS and COVID-19

In this thesis, we develop new mathematical models and data-driven formalisms to address two key questions in infectious disease biology and epidemiology, one pertaining to HIV/AIDS and the other to COVID-19.

Current antiretroviral therapies for HIV infection are not curative. Extensive efforts are ongoing to devise new strategies for eliciting cure. Recently, in a major advance, combinations of drugs called latency reversal agents (LRAs) and broadly neutralizing antibodies (bNAbs) have shown remarkable success in eliciting long-term control of HIV post treatment. How the combinations work when neither LRAs nor bNAbs succeed independently remains unknown. Here, we posit a mechanism of synergy between the two drug classes that may underlie their success in combination. LRAs reactivate latently infected cells, triggering bursts of viral production. bNAbs bind to these virions and enhance antigen presentation, stimulating HIV-specific CD8 T-cells. The combination, administered towards the end of ART, thus leaves a reduced latent reservoir and a primed CD8 T-cell population, which upon ART cessation can lead to long-term control. To elucidate the nature of this control, we advanced mathematical models of within-host HIV dynamics by incorporating the pleiotropic effects of LRAs and bNAbs. The model predicted the existence of two steady states, one with high viral load, marking progressive disease, and the other with low viral load, representing lasting control. Further, with LRAs or bNAbs alone, the latter state was rarely accessed. With the combination, the state was far more frequently realized, explaining the success of the combination. Creating a large virtual patient population mimicking inter-individual variability, we found that model predictions recapitulated data from multiple in vivo studies. The model offered insights into the role of CD8 T-cells in eliciting and maintaining viremic control and presented a route to identifying optimal dosages of LRAs and bNAbs for achieving this control, informing ongoing clinical trials.    

The burden of COVID-19 appeared to vary significantly across nations. Accurately estimating this burden is important for pandemic preparedness. Metrics commonly used to quantify this burden suffer biases. For instance, COVID-19 mortality is affected by healthcare access. Here, we recognised that the prevalence of asymptomatic SARS-CoV-2 infection was unaffected by these biases. Asymptomatic infections reflect basal immunity and occur independently of healthcare infrastructure.  and its variation across nations, however, has not been well studied. We investigated this by conducting a comprehensive analysis of serosurveys from various countries before the widespread availability of vaccines and the emergence of new variants. The studies that met our selection criteria together sampled 4,27,205 individuals and yielded estimates of 28 nations. Using random-effects modelling, we found the pooled global  to be 42.1% (95% CI: 30.0%-55.4%). varied widely across nations (range: 6%-96%), highlighting the enormous underlying variation in the natural immunity to SARS-CoV-2.  Performing meta-regression with national-level metrics, we found that the human development index (HDI) was negatively correlated with  (p=10−4; R2=38.9%). More developed nations thus experienced less frequent asymptomatic infections on average. We speculate on the origins of this unexpected correlation. These findings have implications for unravelling the origins of asymptomatic infections and for future pandemic preparedness.