Integrating molecular engineering and artificial intelligence to overcome clinical and manufacturing challenges in CAR-T therapy
TL;DR
This review presents an integrated perspective on molecular-level CAR-T cell engineering and AI-based computational and manufacturing systems, rather than treating them as separate domains, and examines the role of AI in both developing and processing cell-based CAR-T development of molecular constructs.
Abstract
Chimeric antigen receptor (CAR) T-cell therapy has demonstrated remarkable efficacy in addressing hematological malignancies. However, its broader clinical application in other cancer types is still limited by key challenges like antigen escape, on-target/off-tumor toxicity, reduced efficacy for solid malignancies, and large variability in the production process for CAR T-cells. Traditional optimization approaches are increasingly insufficient to address these complex challenges. Notably, current studies do not fully integrate AI with molecular design and manufacturing processes in CAR-T therapy. Recent advances in the field of artificial intelligence (AI) offer promising solutions to these challenges in an innovative and rational CAR-T cell design using predictive and data-driven reasoning. Machine learning and deep learning models enable the optimization of antigen, scFv affinity and specificity, costimulatory domains, and multiplexed gene editing approaches for the reduction of toxicity and enhancement of therapeutic efficacy. Additionally, the use of AI digital twins and smart bioreactors is transforming CAR-T cell manufacturing into a more controlled, scalable, and reproducible process through real-time monitoring and process optimization. In this review, we present an integrated perspective on molecular-level CAR-T cell engineering and AI-based computational and manufacturing systems, rather than treating them as separate domains. It examines the role of AI in both developing and processing cell-based CAR-T development of molecular constructs, as well as in translating them into a process-based and machine-based pathway. We also analyze key challenges, including limited availability of high-quality datasets, model transparency, interpretability, and constraints in preclinical validation. Based on current advancements in immune cell engineering and AI platforms, we propose a conceptual framework to improve long-term safety, efficacy, and scalability of CAR-T therapies.
