肿瘤细胞治疗中建模方法与机遇的 CAR-T 图谱

Chimeric Antigen Receptor CAR-T therapy represents a potent adoptive cell immunotherapy for cancer, yet its clinical application remains constrained by pronounced interindividual variability, severe adverse events, and restricted efficacy in solid tumors. Computational modeling has emerged as a critical framework for analyzing and characterizing CAR-T cell behavior to mitigate these clinical limitations. This systematic review synthesizes 20 computational models identified through targeted bibliographic search that were developed using human clinical datasets. The reviewed studies employ a range of mathematical frameworks, including cellular kinetics and population pharmacometrics models, tumor-immune interaction models, quantitative systems pharmacology approaches, and multiscale mechanistic models.

Across these frameworks, CAR-T cells are represented as dynamic populations undergoing expansion, contraction, persistence, and functional exhaustion. Many models further incorporate phenotypic stratification into functional subsets, most commonly effector and memory cells, to capture the multiphasic kinetics observed in clinical settings.

Additional variables frequently include tumor burden, antigen expression, host immune cells, cytokines, and CAR-target complexes, reflecting different levels of biological detail and modeling objectives. Based on this analysis, I propose a unified set of core variables that captures the key biological processes represented across existing models while providing a consistent structure for future modeling efforts.

Together, these studies demonstrate that the choice and structure of variables used to describe CAR-T cell populations and their interactions are key determinants of model interpretability and translational relevance. Improved access to longitudinal clinical datasets and phenotype-resolved measurements will be essential for developing more predictive and clinically applicable computational models of CAR-T therapy.