免疫肿瘤学中的空间革命：人工智能解码 NK 细胞生态位以预测治疗反应

Natural killer (NK) cells are important effector cells of the innate immune system and have been investigated as a therapeutic platform for cancer immunotherapy. Although NK cell-based therapies have shown clinical activity in hematological malignancies, their application in solid tumors remains limited by restricted tumor infiltration, functional suppression, and the complexity of the tumor microenvironment (TME). Recent advances in spatial transcriptomics and artificial intelligence (AI) have provided new approaches for characterizing NK cell distribution, functional states, and cellular interactions within the TME. Critically, the spatial distribution and structural organization of NK cells within tumor niches - including their proximity to tumor cells, stromal barriers, and immune effector partners - are fundamental determinants of their cytotoxic function.

A deeper understanding of how spatial context shapes therapeutic response is therefore central to advancing NK cell immunotherapy. This review summarizes how AI-assisted analysis of spatial and multi-omics data may contribute to the discovery and validation of biomarkers associated with NK cell therapy response.

It also discusses the potential applications of AI in optimizing chimeric antigen receptor (CAR)-NK cell engineering, combination therapy strategies, and individualized dosing regimens. By synthesizing studies published in recent years, this review highlights the emerging shift from response prediction toward treatment optimization, while emphasizing the current limitations of available evidence, including model interpretability, data heterogeneity, causal inference, and clinical validation.

Finally, we discuss how four-dimensional (4D) dynamic monitoring and explainable AI may support the future development of more precise and personalized NK cell immunotherapy strategies.