When and how should we intervene to manage an emerging infectious disease most effectively? Deciding when to enforce or relax non-pharmaceutical interventions (NPIs) based on real-time outbreak surveillance data is a central challenge in infectious disease epidemiology. Practical surveillance data, often characterised by reporting delays and infection under-ascertainment, can misinform decision-making. This may lead to mistimed NPIs that fail to control disease spread or allow harmful epidemic peaks that overwhelm healthcare capacities.
In this talk, I will introduce EpiControl, a novel model-predictive control algorithm designed to optimise NPI decisions by jointly minimising cumulative future risks and costs across stochastic epidemic projections. I will demonstrate how this algorithm outperforms data-insensitive strategies while also discussing the intrinsic limitations of surveillance quality, disease growth rates, and decision frequency in flattening epidemic peaks or reducing endemic oscillations. Additionally, I will present my ongoing research on integrating population behaviour into the policy-making framework.
Inhalational anthrax, caused by the bacterium Bacillus anthracis, is a disease with very high fatality rates. Due to the significant risk posed if the bacterium was to be intentionally used as a bioweapon, it is important to be able to defend against such an attack and to make optimal decisions about treatment strategies. Mechanistic mathematical models can help to quantify and improve understanding of the underlying mechanisms of the infection. In this talk, I will present a multi-scale mathematical model for the infection dynamics of inhalational anthrax. This approach involves constructing individual models for the intracellular, within-host, and population-level infection dynamics, to define key quantities characterising infection at each level, which can be used to link dynamics across scales. I will begin by introducing a model for the intracellular infection dynamics of B. anthracis, which describes the interaction between B. anthracis spores and host cells. The model can be used to predict the distribution of outcomes from this host-pathogen interaction. For example, it can be used to estimate the number of bacteria released upon rupture of an infected phagocyte, as well as the timing of phagocyte rupture and bacterial release. Next, I will show how these key outputs can be used to connect the intracellular model to a model of the infection at the within-host scale. The within-host model aims to provide an overall understanding of the early progression of the infection, and is parametrised with infection data from studies of rabbits and guinea pigs. Furthermore, this model allows the probability of infection and the time to symptoms to be calculated. Building a model that offers a realistic mechanistic description of different animal responses to the inhalation of B. anthracis spores is an important step towards eventually extrapolating the model to describe the dynamics of human infection. This would enable predictions of how many individuals would become infected in different exposure scenarios and also on what timescale this would occur.
When managing a breeding programme, we want to maximize the selection of desirable traits (such as health or yield). At the same time, we know that related plants or animals are more likely to share traits, so we also need to incorporate minimizing inbreeding and its associated risks. This can be modelled as a bi-objective optimization problem, which happens to have a similar structure to portfolio theory from financial mathematics.
Collaborating with researchers at the Roslin Institute in the Royal (Dick) School of Veterinary Studies, we examined how a range of mathematical tools can be used to explore this problem more accurately and efficiently than the state of the art. These were tested with simulated breeding programmes and led to the creation of an open-source tools for practitioners.
At Leeds, we have been developing within-host models of infection for a number of years. Deterministic models of viral dynamics have been widely used in the past to understand average behaviours, and to quantify the efficacy of treatments. On the other hand, stochastic models allow one to incorporate discrete events (such as cellular burst, where an infected cell dies releasing N virions into the extracellular environment), to look at extinction events (e.g. probably of infection establishment vs short-time recovery of the host, for a given initial dose), and to account for variability in summary statistics such as the reproduction number (i.e., the number of cells infected by a typical infected cell during an infection). These heterogeneities can be important for example when looking at extinction events and/or the impact of small viral doses. Stochastic models also allow one to obtain summary statistics related to the infection dynamics across different scales (intra-cellular, within-host and population levels), and to link these scales via multi-scale modelling approaches. In this talk, I will discuss recent work that we have carried out in this area.
Preparing for the next pandemic: development of mathematical and computational tools for the assessment of the impact of novel antivirals at the individual and population level.
Some first modelling efforts towards the assessment of a new approach to combat pathogenic respiratory viruses using novel broadly neutralising antibodies will be presented.
Effective collective decision-making in human and animal groups requires robust mechanisms to form consensus, typically via feedback loops in which individuals adapt their behaviour based on their perception of others. Such behaviour has been observed and theorised across scales from nucleosomes to entire societies. Of equal importance, but far less well studied, is the question of how consensus is overturned. In many contexts it is vital that group decisions do not remain fixed in the face of new evidence; echo-chamber effects must be suppressed so that the collective preferences which are expressed are not too strongly entrenched. In this talk I will discuss a new mathematical theory for how consensus can be overturned in symmetric binary choice problems, and compare the theoretical predictions to experiments with human and animal groups.
West Nile virus (WNV) is a vector-borne pathogen causing major outbreaks of West Nile fever worldwide. Although the transmission is maintained via birds and mosquitoes, human infection is possible. Routine surveillance of WNV in the USA is performed by trapping mosquitos and testing for the presence of WNV during the transmission season by RT-qPCR testing. Apart from the general binary positive/negative outcome from these tests, they also generate cycle threshold (Ct) values. Motivated by findings from SARS-CoV-2 viral load dynamics in humans, we hypothesised that Ct values observed through routine pooled testing over time are sufficient to provide information on WNV epidemic dynamics. To investigate this, we introduce an agent-based model of mosquitoes and birds to model WNV epidemiological and viral load dynamics. In this model, we embed a within-host model of the viral load kinetics of mosquitoes. We simulate scenarios where mosquitoes are captured through routine surveillance and tested for WNV through pooled RT-qPCR testing, generating synthetic Ct values over time. We compare our model output to real Ct value data collected through WNV pooled testing in Nebraska in 2022 and 2023.
