Integrating network toxicology, machine learning, and single-cell sequencing systems to analyze autophagy core genes...

The heterogeneity and complex tumor microenvironment of lung adenocarcinoma lead to poor prognosis. Autophagy, as a key cellular process, interacts with tumor immune infiltration and jointly affects the progression of lung adenocarcinoma, but its core regulatory genes and mechanisms are still unclear.This study integrated three lung adenocarcinoma transcriptome datasets from the GEO database and performed cross-analysis with the human autophagy gene set to screen for differentially expressed autophagy-related genes. Identify core prognostic genes by constructing protein interaction networks and combining them with machine learning algorithms (Cox regression, SHAP analysis). Use the independent dataset GSE68465 to validate the model through a combination of 100 algorithms. Further elucidate the function, cellular localization, and association with smoking risk of core genes through enrichment analysis, immune infiltration assessment, single-cell transcriptome analysis, and network toxicology.A total of 276 shared autophagy‑related differentially expressed genes were identified. Using machine learning, five core genes-ENG, CDH1, KLF4, IL6, and MMP9-were selected. In the independent validation cohort, the prognostic model based on these genes demonstrated strong diagnostic performance (AUC > 0.9). Enrichment analysis revealed that the core genes were significantly enriched in pathways such as cellular senescence, autophagy, and FoxO signaling. Immune infiltration analysis showed that M1 macrophages and naïve B cells were significantly upregulated in tumor tissues, whereas resting dendritic cells were downregulated. Single‑cell analysis identified specific expression of genes including CDH1 and TGFB1 in type II alveolar cells and immune cells. Network toxicology and molecular docking confirmed that nicotine, a major component of cigarette smoke, exhibits high‑affinity binding to the core genes CDH1, HIF1A, KLF4, TGFB1, and BCL2.This study successfully identified and validated a robust prognostic feature consisting of five autophagy-related genes. These genes play a key role in the development of lung adenocarcinoma by regulating the tumor immune microenvironment, cell communication, and responding to external risk factors, providing new potential targets for prognosis prediction and targeted therapy.