将网络分析与机器学习相结合，探索齐考昔酸在肺腺癌中的药理学靶点及相关作用机制

Abstract

Background Lung adenocarcinoma (LUAD), a major type of non-small cell lung cancer, has high incidence and mortality rates. Oleanolic acid (OA), a natural pentacyclic triterpene compound, has demonstrated potential anti-tumor effects but its pharmacological effects in LUAD remain unclear. Purpose This study aims to identify the potential targets and action pathways of OA in LUAD using network analysis, transcriptomics, molecular docking, animal experiments, and preliminary clinical sample assessment. Methods Potential candidate genes for OA in LUAD were identified from public databases including PubChem, GeneCards, and TCGA. GO and KEGG enrichment analyses were performed using Metascape, and a PPI network was constructed. Machine learning methods (Lasso regression and SVM-RFE) were employed to refine the candidate genes and identify key biomarkers. Prognostic, clinicopathological, GSEA, mutation, and immune infiltration analyses were conducted. Molecular docking was performed to estimate the theoretical binding affinity between OA and the biomarkers, which was further evaluated using animal models and clinical samples. Results 21 candidate genes were initially identified, with TOP2A and ALOX5AP emerging as key biomarkers. High TOP2A expression and low ALOX5AP expression were associated with poor prognosis. GSEA showed co-enrichment in the IgA production pathway. Mutation analysis indicated higher amplification propensity for TOP2A. Both biomarkers correlated with immune cells. Molecular docking and animal experiments suggested OA’s regulatory effects on these genes, with clinical sample evaluation. Conclusion Utilizing computational screening as a theoretical starting point alongside preliminary experimental evaluations, this study identifies ALOX5AP and TOP2A as potential pharmacological biomarkers associated with OA treatment in LUAD. Rather than asserting definitive specific targeting, these findings provide preliminary associative evidence. Although direct