📝 SummarXiv

Abstract

Chimeric antigen receptor (CAR) T cell therapy has shown remarkable success in hematological malignancies, yet patient responses remain highly variable and the roles of CD4 and CD8 subsets are not fully understood. We present an extended mathematical framework of CAR-T cell dynamics that explicitly models CD4 helper and CD8 cytotoxic lineages and their interactions with tumor antigen burden. Building on the Kirouac et al. (2023) model of antigen-regulated memory, effector, and exhaustion transitions, our system of differential equations incorporates cytokine-mediated modulation of CD8 proliferation, cytotoxicity, and memory regeneration by CD4 T cells. Sensitivity analyses identify effector proliferation burst size, antigen turnover, and CD8 expansion rates as dominant determinants of treatment outcome. Virtual patient simulations reproduce clinical findings that a 1:1 CD4:CD8 ratio enhances CAR-T expansion and tumor clearance relative to CD8-only products. Finally, we integrate a feed-forward neural network trained on noisy virtual patient data to improve predictive robustness, and apply SHAP analysis to interpret the network predictions and compare them with mechanistic sensitivity analyses. This work highlights the synergistic roles of CD4 and CD8 CAR-T cells and provides a quantitative foundation for optimizing treatments and patient stratification.