Constructing a novel prognostic model of gastric cancer related to mitochondrial membrane permeability change-driven necrosis based on single cell and bulk transcriptome

Background Gastric cancer (GC) is the fifth most common cancer in the world, with a late diagnosis and poor prognosis. The mitochondrial permeability transition-driven necrosis (MPTDN) is often associated with cancer, while its mechanism in GC is unclear. Methods A single-cell RNA sequencing (scRNA-seq) dataset GSE183904, and two mRNA profile datasets TCGA-stomach adenocarcinoma (STAD) and GSE62254 were downloaded from the online databases. After a series of analyses of GSE183904, the differentially expressed genes (DEGs) in different proportions of cells (DCs) were selected between GC and controls for further analysis, namely DC-DEGs. The DEGs between STAD and normal controls (STAD-DEGs) were screened, and then highly correlated modules with MPTDN-related genes (MPTDN-RGs) were obtained by Weighted gene co-expression network analysis (WGCNA). Next, DC-DEGs, STAD-DEGs, and module genes were overlapped as candidate genes. The prognostic genes were selected via the Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression model, and the prognostic risk model was constructed and verified. Then, the immune cell infiltrations and Immunotherapy response were conducted. Later, pseudotime analyses were performed to explore key cell evolution trajectories. Furthermore, the clinic specimens were collected to perform qPCR analysis. Results Epithelial cells and Tissue stem cells were obtained as DCs, and 1,592 DC-DEGs were identified between GC and normal controls. After the overlapping of 2,238 STAD-DEGs, 1,592 DC-DEGs, and 3,832 module genes, a total of 112 candidate genes were determined. Then, 4 prognostic genes ( GPX3 , CD36 , VCAN , and SERPINE1 ) were identified the risk model was constructed on this basis. Afterward, risk score, age, and N categories were screened as the independent prognostic factors to construct a nomogram model, which could effectively predict the 1-, 3-, and 5-year OS of STAD patients. Subsequently, Endothelial cells and Tissue stem cells differentiated in normal control were less than those of GC. Finally, Moreover, the qPCR revealed that the expression of GPX3 and CD36 was reduced in GC tumor tissues. Conclusion Overall GPX3 , CD36 , VCAN , and SERPINE1 could be potential therapeutic targets to diagnose and treat GC. Tissue stem cells were key cells in GC patients and provided strong support for further understanding of GC.