Abstract
Natural killer (NK) cells are an integral component of the tumor microenvironment, and their role in immune checkpoint inhibitors (ICI) therapy has garnered increasing attention. However, comprehensive studies on NK cells across cancers, especially their impact on immunotherapy response, remain limited. We used machine learning algorithms to establish a pan-cancer natural killer cell immunotherapy predictive model (NKCIPM) by combining single-cell RNA sequencing data from 164 samples across 6 cancer types and bulk RNA-seq data from different tumor samples. Tumor immune cell infiltration analysis, drug sensitivity analysis, and cell-cell communication were also further conducted. An upregulation of NK cell proportions post-immunotherapy and the identification of 188 NK cell differentially expressed genes were observed through single-cell RNA sequencing analysis. By integrating bulk RNA-seq data and applying machine learning algorithms, 7 key hub genes were identified, ultimately leading to the construction of NKCIPM, with APOE emerging as the most influential hub gene. Further analysis using the CIBERSORT algorithm revealed that the signature genes within this model were significantly associated with immune cell infiltration and response to ICI. Additionally, therapeutic evaluation of CHEK1 and CHEK2 targets demonstrated potential significance in the communication between B cells, NK cells, and mast cells within the context of ICI therapy. In summary, the NKCIPM model offers a valuable tool for predicting immunotherapy outcomes and informing clinical decision-making, highlighting the potential of NK cell signature genes as therapeutic targets.