Unleashing the potential role of tumor-associated NK cells as a novel immunotherapeutic target in triple-negative breast cancer.

Breast cancer is the most prevalent malignancy among women worldwide, with triple-negative breast cancer (TNBC) comprising 10–20% of cases. Although immunotherapy has recently emerged as a promising treatment, its efficacy remains limited in immunodeficient subtypes such as TNBC. Tumor-associated NK cells (TaNKs), an immunologically exhausted NK cell subset, have not yet been characterized in TNBC. Elucidating the TNBC tumor microenvironment through TaNKs may therefore provide novel diagnostic and therapeutic insights. Here, we employed single-cell transcriptomics to characterize TaNKs, which are broadly distributed across cancers. TaNKs displayed profound functional impairment, with diminished cytotoxicity and heightened stress responses, thereby compromising immune surveillance. Bulk RNA-seq analysis revealed that elevated TaNKs abundance correlated with poor prognosis in TNBC patients. Using machine learning, we established a TaNKs feature score (TaNKFS) incorporating six genes (HSPA1B, TUBB2A, BAG3, NR4A2, IER2, and MYADM), which demonstrated strong prognostic value and effectively quantified TaNKs levels in TNBC. Notably, immune checkpoint inhibitors showed minimal benefit in cases with high TaNKFS. Furthermore, deep learning-based drug screening and high-throughput molecular docking identified twelve candidate compounds targeting TaNKFS, including six clinically available agents (Staurosporine, Bexarotene, Cyclophosphamide, Vancomycin, Heroin, and Fluorouracil). Synergy prediction further suggested multiple effective combination regimens. In summary, this study delineates the role of TaNKs in TNBC, establishes a prognostic biomarker, and proposes novel therapeutic strategies integrating machine learning and deep learning approaches.