(T1130-05-27) Characterizing Population Variability in Vaccine-induced Immunity: A Machine Learning Integrated Mechanistic Modeling Approach

Purpose: Understanding the heterogeneous nature of vaccine-induced immune responses is critical for optimizing immunization strategies across diverse populations. While mechanistic models offer insight into immunological processes, they are often limited in scalability across large clinical datasets. To address this, we present a machine learning–integrated mechanistic modeling framework to quantitatively characterize population variability in vaccine response and identify key immunological parameters underlying differential immunity.

Methods: We developed a mechanistic model of immune response to mRNA vaccination, describing antigen kinetics, CD4+ T cell activation, plasma cell differentiation, cytokine signaling, and antibody production. The model comprises a system of ordinary differential equations and was fit to longitudinal antibody titer data from ~3,500 individuals over 230 days post-vaccination. We estimated key immunological parameters using subject-specific model fitting, followed by global sensitivity analysis to identify dominant determinants of antibody response. Clustering analysis of antibody kinetics was performed to identify immune response subtypes and correlated with model-fitted parameters. Finally, we trained machine learning models to predict key immunological parameters based on demographic features and early immune responses as a tool for prospective predictions and optimization simulations.

Results: Model fitting demonstrated strong concordance with clinical data (median R > 0.9 across individuals). Sensitivity analysis revealed antibody production rate per cell and plasma cell activation rate as key drivers of response heterogeneity. Clustering identified distinct responder subgroups: high, moderate, and low responders, with statistically significant differences in model parameters. ML models accurately predicted these parameters from demographic and early-response data (R² > 0.8), enabling prospective identification of immune response phenotypes.

Conclusion: This integrative approach enables mechanistic interpretation of vaccine response variability at the population level while leveraging machine learning for scalable parameter inference. The framework facilitates immune fingerprinting and personalized prediction of vaccine efficacy, supporting data-driven strategies for dose scheduling, trial design, and precision vaccination.

References: 1. Klein SL, Flanagan KL. Sex differences in immune responses. Nature Reviews Immunology. 2016/10/01 2016;16(10):626-638. doi:10.1038/nri.2016.90
2. Lynn DJ, Benson SC, Lynn MA, Pulendran B. Modulation of immune responses to vaccination by the microbiota: implications and potential mechanisms. Nature Reviews Immunology. 2022/01/01 2022;22(1):33-46. doi:10.1038/s41577-021-00554-7
3. Dogra P, Schiavone C, Wang Z, et al. A modeling-based approach to optimize COVID-19 vaccine dosing schedules for improved protection. JCI Insight. 05/25/ 2023;doi:10.1172/jci.insight.169860

Acknowledgements: This work was supported in part by the National Institutes of Health (NIH) Grant 1R01EB035545 (P.D.).