Single-cell and multi-omics analysis reveals the role of stem cells in prognosis and immunotherapy of lung adenocarcinoma patients

Jianan Zheng¹Haoran Lin^1,2Wei Ye¹Mingjun Du^1*Chenjun Huang^1*Jun Fan^1*

• ¹The First Affiliated Hospital With Nanjing Medical University, Nanjing, China
• ²Nanjing Medical University, Nanjing, China

The roles of stem cells in lung adenocarcinoma (LUAD) progression and therapeutic resistance have been recognized, yet their impact on patient prognosis and immunotherapy response remains unclear. Methods: Single-cell RNA sequencing was performed to identify stem cell populations characterized by high expression of MKI67 and STMN1. Key marker genes were identified using the FindAllMarkers function, and these genes were subsequently analyzed for mutations, copy number variations, and prognostic significance in LUAD patients. Multiple machine learning algorithms were systematically compared in order to develop an optimal prognostic model. The predictive performance of the model was validated across seven independent LUAD cohorts and immunotherapy datasets. Patterns of immune infiltration were assessed using various computational approaches and were further validated in an internal hospital cohort. Results: Through comprehensive machine learning optimization, CoxBoost+Enet (alpha=0.7) was identified as the optimal model, incorporating seven key stem cell-related genes and designated as the Stem Cell Prognostic Model (SCPM). Patients were consistently stratified into high-and low-SCPM groups across all seven validation cohorts, with poorer overall survival observed in the high-SCPM group. Predictive accuracy was demonstrated by ROC analysis (AUC > 0.65), while clear group separation was confirmed through PCA based on the seven-gene signature. Notably, immunotherapy response was also predicted by SCPM, with inferior outcomes observed in high-SCPM patients following treatment with immune checkpoint inhibitors. Significantly lower immune cell infiltration, characteristic of "cold" tumors, was detected in high-SCPM patients by multiple immune infiltration algorithms. These findings were further validated in the internal cohort, where reduced CD8+ T cell infiltration was observed in high-SCPM patients. Conclusion: A stem cell-based prognostic model (SCPM) was constructed and validated, enabling accurate prediction of survival and immunotherapy response in LUAD patients. Patients with immunologically "cold" tumors, as identified by the SCPM, may benefit from alternative therapeutic strategies.