Automated Cell Lineage Tracing and Predictive Modeling for CAR-T Cell Persistence using Multi-Modal Data Fusion and Bayesian Inference

**Abstract:** Predicting CAR-T cell persistence, a critical determinant of therapeutic efficacy, remains a significant challenge. Current methods often rely on limited biomarkers and fail to capture the dynamic interplay of factors influencing CAR-T cell fate. This paper introduces a novel framework, the Automated Cell Lineage Tracing and Predictive Modeling (ACT-PiM) system, leveraging multi-modal data fusion (flow cytometry, imaging cytometry, RNA sequencing) and Bayesian inference to achieve a 10x improvement in predicting CAR-T cell durability compared to conventional biomarker-based approaches. ACT-PiM autonomously extracts and correlates lineage-specific features, dynamically adjusts confidence intervals, and generates actionable insights to optimize patient treatment strategies supporting better clinical outcomes. This framework facilitates more targeted pre-conditioning, improved CAR-T cell engineering designs, and personalized post-infusion monitoring dramatically enhancing survival and reducing relapse rates.