Abstract
BACKGROUND: Lung adenocarcinoma (LUAD) is characterized by high mortality and a complex pathogenesis. Despite significant advancements in targeted therapies and immunotherapies, patient prognosis remains poor, and there is a notable absence of effective biomarkers for early diagnosis and treatment. This study employs bioinformatics methods to identify key genes with diagnostic and therapeutic potential in LUAD and constructs a predictive model based on these genes. METHODS: Genetic expression data and clinical information from LUAD patients and healthy controls were obtained from the GWAS and TCGA databases. Differentially expressed genes (DEGs) and disease-related genes were identified through Mendelian randomization (MR) analysis. GO functional enrichment and KEGG pathway analyses were performed, followed by Cox regression and LASSO regression to identify potential diagnostic and therapeutic target genes. Kaplan-Meier survival curves and a Line chart were generated to predict 1-, 3-, and 5-year survival rates. RESULTS: MR analysis identified 16 genes related to LUAD development, including WFDC3, FANCD2, OTX1, and others. Cox and LASSO regression pinpointed three significant genes: CD33, FANCD2, and TFPI. Kaplan-Meier curves showed higher survival rates for low-risk FANCD2 and TFPI groups, while the high-risk CD33 group had elevated survival. The calibration curve in the validation set confirmed the predictive accuracy of the model. CONCLUSION: This study presents a prediction model based on CD33, FANCD2, and TFPI, which could aid in individualized treatment decisions and provide a basis for further LUAD research.