Abstract
Epithelial cells are integral to tumor composition and engage with various immune cell types within the tumor microenvironment, influencing tumor progression and metastasis. A thorough exploration of the roles and mechanisms of these epithelial cells could enhance early detection strategies and treatment modalities for lung adenocarcinoma (LUAD). This research employed single-cell analysis techniques, complemented by machine learning algorithms, to identify genes associated with epithelial cells and evaluate their prognostic significance and implications for immunotherapy in LUAD patients. By leveraging multiple datasets and applying diverse clustering methods within machine learning, we successfully crafted and validated a diagnostic model specifically for LUAD. Among the genes linked to epithelial cells, the XGBoost and random forest techniques identified PMAIP1 as the most crucial gene in terms of prognosis. Additionally, this study investigated the relationship between PMAIP1 and the infiltration of immune cells. The expression levels of PMAIP1 and its relevance in LUAD were subsequently confirmed through immunohistochemical staining and in vitro cell experiments. This analysis revealed 17 key genes associated with epithelial cells by integrating single-cell analysis with clinical data from the TCGA-LUAD dataset, underscoring their significance in diagnosis, prognostic assessment, and possible treatment avenues for LUAD patients. Importantly, PMAIP1 is strongly linked to prognosis and responses to immunotherapy in LUAD, with experimental findings indicating its heightened expression in PRAD and its connection to adverse outcomes. Furthermore, reducing PMAIP1 expression has been shown to hinder the proliferation, metastasis, and stemness of LUAD cells. In summary, our findings indicate that PMAIP1 has potential as a prognostic biomarker and a target for immunotherapy in patients with LUAD.