A Wnt-related gene expression signature to improve the prediction of prognosis and tumor microenvironment in gastric cancer

Background: Most gastric cancer (GC) patients were diagnosed in the advanced stages without obvious symptoms, which resulted in the increased risk of death. Although the combination therapies have showed survival benefit of patients, there is still urgent need to explore the underlying mechanisms of GC development and potential novel targets for clinical applications. Numerous studies have reported the upregulation of Wnt signaling pathway in human GC, which play important role during GC development and progression. However, the current understanding of Wnt signaling pathway is still limited due to its complexity and contradictory effect on different stages of GC tumor microenvironment. Method: We used The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) dataset to screen Wnt signaling pathway-associated genes by ssGSEA and correlation analysis. Three molecular subtypes were constructed based on a consistent clustering analysis. The key Wnt-related genes were screened through univariate cox analysis, lasso, and stepwise regression. In addition, the Gene Set Enrichment Analysis (GSEA) were performed to explore potential molecular pathways regulated by the Wnt-related gene signatures. ESTIMATE was utilized for evaluating the immune cell populations in GC tumor microenvironment. Results: Three molecular subtypes associated to Wnt were identified, and 7 key Wnt-related genes were screened to establish a predictive RiskScore model. These three molecular subtypes showed significant prognostic differences and distinct functional signaling pathways. We also found the downregulated immune checkpoint expression in the clust1 with good prognosis. The RiskScore model was successfully validated in GSE26942 dataset. Nomogram based on RiskScore and Gender had better prognostic predictive ability. Conclusion: In summary, our study showed that the Wnt-related genes could be used to predict prognosis of GC patients. The risk model we established showed high accuracy and survival prediction capability.