Investigation Into the Association Between Neurotransmitters, Immune Features, and Lung Adenocarcinoma: Identifying GABA-Related Features Using Machine Learning Methods.

Background: Lung adenocarcinoma (LUAD), a predominant subtype of non-small cell lung cancer (NSCLC), is associated with a high mortality rate. Currently, there are no reliable or sensitive biomarkers or prognostic methodologies available for its early detection or diagnosis. Gamma-aminobutyric acid (GABA), a pivotal inhibitory neurotransmitter within the central nervous system (CNS), primarily exerts its effects through interactions with GABA receptors (GABARs). Recent studies have increasingly highlighted GABA's significant role in mediating the initiation and progression of various tumors beyond the CNS. Nonetheless, research investigating the role of GABA in LUAD is limited, and the specific molecular and cellular mechanisms underlying its interactions remain to be fully elucidated. Methods: We developed an innovative machine learning framework designed to screen GABA-related genes (GABARgenes) at both single-cell and large transcriptomic levels. This framework encompasses 10 algorithms and 101 combinatorial pairing patterns, which facilitate the construction of consistent GABA-related features (GABARFs). The framework's performance was assessed using both a training set and an external validation set. To provide a quantitative prognostic tool for clinical application, we established a nomogram that incorporates GABARF. Additionally, we conducted multiomics analyses, including genomics, single-cell transcriptomics, and comprehensive transcriptomics, to derive and consolidate more extensive prognostic features. We also evaluated the response of GABARF-defined risk subgroups to immunotherapy and identified potential personalized therapeutic agents for specific risk categories. Results: Among the 124 GABARgenes analyzed, 38 demonstrated a significant correlation with overall survival (OS) in patients. Our machine learning-derived GABARF exhibited exceptional performance in predicting prognosis and clinical outcomes, showing promise in forecasting the onset and progression of LUAD. Multivariate analysis confirmed that GABARF serves as an independent prognostic factor for OS in LUAD. Furthermore, distinct GABARF risk subgroups exhibited significant differences in biological function, mutation status, and tumor immune infiltration. Notably, there were significant variations in the immunophenoscore (IPS) across the risk subgroups. GABARF risk stratification aligns with stemness properties of tumor cells, indicating that high-risk patients may harbor tumors with enhanced stemness features that contribute to their poor prognosis and reduced immunotherapy response. Sensitivity analyses of conventional LUAD therapies indicated that patients in the low-risk group may derive greater benefit from immune checkpoint inhibitors (ICIs), while those in the high-risk group may exhibit heightened sensitivity to first-line chemotherapy agents. Furthermore, LDHA overexpression was found to promote proliferation and migration, while inhibiting apoptosis. In addition, overexpression of LDHA can upregulate the expression of stemness markers CD133, SOX2, and OCT4 in LUAD cells, enhancing the malignant phenotype of tumor cells. Conclusion: This study presents a novel machine learning-based model for GABARF, which shows promise as a potential tool to aid in prognostic prediction, targeted prevention, and individualized treatment planning in LUAD. Initial investigations into the interaction mechanisms of GABARF at the molecular, cellular, and tumor immune microenvironment (TIME) levels in LUAD have commenced. The GABARF model not only serves as a prognostic indicator but may also reflect the stemness status of LUAD tumors, offering insights into personalized treatment strategies that account for both neural-immune-stemness interactions.