Designing a risk prognosis model based on natural killer cell-linked genes to accurately evaluate the prognosis of gastric cancer

Abstract

Background: This study was aimed at identifying natural killer (NK) cell-related genes to design a risk prognosis model for the accurate evaluation of gastric cancer (GC) prognosis. Methods: We obtained NK cell-related genes from various databases, followed by Cox regression analysis and molecular typing to identify prognostic genes. Various immune algorithms and enrichment analyses were used to investigate the mutations, immune status, and pathway variations among different genotypes. The key prognostic genes were assessed using the least absolute shrinkage and selection operator (Lasso) regression analysis and univariate Cox regression analysis. Thereafter, the risk score (RS) prognosis model was constructed based on the selected important prognostic genes. A Receiver Operating Characteristics (ROC) curve was plotted for analyzing the robustness of the model. Subsequently, the decision and calibration curves were used for assessing the reliability and prediction accuracy of the proposed model. The ‘pRRophetic’ R software package was utilized for predicting the half-maximal inhibitory concentration (IC50) of immunotherapy and chemotherapy drugs. Results: We screened 21 prognostic genes and three molecular subtypes and found that the C1 subtype had the worst prognosis. Further, the pathways promoting tumor proliferation, such as epithelial-mesenchymal transition were significantly up-regulated. The results also showed that the macrophages in the M2 stage were significantly infiltrated in the C1 subtype, and there was significant overexpression in the C1 subtype, accompanied by a severe inflammatory reaction. The C1 was highly sensitive to drugs like 5-fluorouracil and paclitaxel. The ROC, calibration curve, and decision curve showed that the risk model was robust and strongly reliable. Conclusion: Overall, our proposed NK cell-related RS model can be used as a more accurate prediction index for GC patients, providing a valuable contribution to personalized medicine.