Machine learning developed a programmed cell death signature for predicting prognosis and immunotherapy benefits in lung adenocarcinoma

Abstract

BACKGROUND: Lung cancer is the leading cause of cancer-related deaths worldwide with poor prognosis. Programmed cell death (PCD) plays a crucial function in tumor progression and immunotherapy response in lung adenocarcinoma (LUAD). METHODS: Integrative machine learning procedure including 10 methods was performed to develop a prognostic cell death signature (CDS) using TCGA, GSE30129, GSE31210, GSE37745, GSE42127, GSE50081, GSE68467, GSE68571, and GSE72094 dataset. The correlation between CDS and tumor immune microenvironment was evaluated using various methods and single cell analysis. qRT-PCR and CCK-8 assay were conducted to explore the biological functions of hub gene. RESULTS: The prognostic CDS developed by Lasso + survivalSVM method was regarded as the optimal prognostic model. The CDS had a stable and powerful performance in predicting the clinical outcome of LUAD and served as an independent risk factor in TCGA and 8 GEO datasets. The C-index of CDS was higher than that of clinical stage and many developed signatures for LUAD. LUAD patients with low CDS score had a higher PD1&CTLA4 immunophenoscore, higher TMB score, lower TIDE score and lower tumor escape score, indicating a better immunotherapy benefit. Single cell analysis revealed a strong and frequent communication between epithelial cells and cancer-related fibroblasts by specific ligand-receptor pairs, including COL1A2-SDC4 and COL1A2-SDC1. Vitro experiment showed that SLC7A5 was upregulated in LUAD and knockdown of SLC7A5 obviously suppressed tumor cell proliferation. CONCLUSION: Our study developed a novel CDS for LUAD. The CDS served as an indicator for predicting the prognosis and immunotherapy benefits of LAUD patients.