From antibody kinetics to correlates of protection : insights from within-host mathematical models

Source abstract

Despite major advances in therapeutics and vaccination, understanding humoral immune responses remains challenging. Antibodies play an important role in protection, yet antibody measurements alone do not capture the complexity of the immune system. Understanding how antibody responses evolve, which biological processes sustain them, and how they relate to protection is important for interpreting immune measurements and informing vaccination strategies. This dissertation therefore explored how within-host mathematical models can be used to describe antibody kinetics and investigate their relationship with protection. First, a systematic review examined mathematical approaches used to model antibody kinetics. Most studies relied on phenomenological models that describe antibody trajectories, while mechanistic models linking these trajectories to underlying biological processes remained relatively scarce. Mechanistic approaches also varied considerably in structure and were strongly constrained by the type and availability of immunological data. Building on these findings, a mechanistic model was applied to longitudinal antibody responses following a two-dose Ebola vaccine regimen and subsequent booster vaccination in healthcare workers in the Democratic Republic of the Congo. Antibody decay was slower following booster vaccination than after the primary regimen. While the model characterized short-term processes underlying antibody responses, long-term mechanisms could not be reliably inferred from antibody measurements alone. This highlighted how biological insight from mechanistic models depends on model structure, identifiability, study design, and the availability of complementary immune markers. The relationship between antibodies and protection was subsequently investigated using SARS-CoV-2 breakthrough infections in vaccinated oncological patients. Longitudinal measurements were used to reconstruct antibody trajectories and estimate antibody levels around the time of exposure. Higher antibody levels were associated with lower risk of symptomatic infection, although this relationship varied across viral variants and host characteristics. Importantly, uncertainty was explicitly incorporated into the estimation of correlates of protection, demonstrating that these estimates depend not only on biology but also on data availability, study design, modelling assumptions, and methodological choices. Overall, this dissertation demonstrates the complementary value of mechanistic and phenomenological modelling for understanding antibody kinetics and protection. Future studies should integrate modelling considerations into study design and combine longitudinal antibody measurements with complementary immune markers to better characterize the biological processes underlying antibody responses and their relationship with protection.