Deciphering Dormant Cells of Lung Adenocarcinoma: Prognostic Insights from O-glycosylation-Related Tumor Dormancy Genes Using Machine Learning.

Lung adenocarcinoma (LUAD) poses significant challenges due to its complex biological characteristics and high recurrence rate. The high recurrence rate of LUAD is closely associated with cellular dormancy, which enhances resistance to chemotherapy and evasion of immune cell destruction. Using single-cell RNA sequencing (scRNA-seq) data from LUAD patients, we categorized the cells into two subclusters: dormant and active cells. Utilizing high-density Weighted Gene Co-expression Network Analysis (hdWGCNA) and pseudo-time cell trajectory, aberrant expression of genes involved in protein O-glycosylation was detected in dormant cells, suggesting a crucial role for O-glycosylation in maintaining the dormant state. Intercellular communication analysis highlighted the interaction between fibroblasts and dormant cells, where the Insulin-like Growth Factor (IGF) signaling pathway regulated by O-glycosylation was crucial. By employing Gene Set Variation Analysis (GSVA) and machine learning, a risk score model was developed using hub genes, which showed high accuracy in determining LUAD prognosis. The model also demonstrated robust performance on the training dataset and excellent predictive capability, providing a reliable basis for predicting patient clinical outcomes. The group with a higher risk score exhibited a propensity for adverse outcomes in the tumor microenvironment (TME) and tumor mutational burden (TMB). Additionally, the 50% inhibitory concentration (IC50) values for chemotherapy exhibited significant variations among the different risk groups. In vitro experiments demonstrated that