Causal Machine Learning Methods to Study Individual Vaccine Efficacy Using Multi-Source Data

The long-term objective of this project is to advance vaccine efficacy (VE) studies by developing novel statistical and machine learning methods that characterize heterogeneity in vaccine responses. Current VE studies primarily focus on population-level averages, obscuring variability across individuals and study contexts. To overcome these limitations, this project will develop robust prediction intervals for individual vaccine efficacy (IVE) using conformal inference, integrate information from multiple VE trials via privacy-preserving federated learning, and develop time-to-event methods to quantify VE that remain valid across study sites, regions, and calendar periods. These methods will accommodate surrogate markers, such as immune correlates of protection, to improve efficiency in treatment effect estimation. The research will leverage data from six harmonized COVID-19 VE trials to develop assumption-lean methods that allow for individual-level predictions while ensuring privacy, thereby enhancing the robustness of VE estimates and improving their applicability across different populations. The candidate will receive comprehensive training in virology, immunology, genomic sieve analysis, immune correlates of protection, and advanced statistical and machine learning techniques. This training will equip the candidate with the skills needed to develop innovative approaches for understanding variability in vaccine responses and to lead future VE research. An interdisciplinary mentorship team spanning biostatistics, infectious disease epidemiology, network science, immunology/virology, and VE trials will support the candidate in achieving these goals. The project will also involve collaborations with leading public health institutions to ensure that the methods developed are directly applicable to real-world challenges. By creating a flexible and scalable framework, this research has the potential to influence the design and analysis of future vaccine trials. Ultimately, the project aims to create a methodological framework that provides more precise VE estimates, informs personalized vaccine strategies, and has a significant impact on public health policy decisions, leading to more effective vaccine deployment and optimization of strategies to combat infectious diseases, to the benefit of all Americans. Modified