A methylation-related signature for predicting prognosis and sensitivity to first-line therapies in gastric cancer.

Abstract

Methylation modification patterns play a crucial role in human cancer progression, especially in gastrointestinal cancers. We aimed to use methylation regulators to classify patients with gastric adenocarcinoma and build a model to predict prognosis, promoting the application of precision medicine. We obtained RNA sequencing data and clinical data from The Cancer Genome Atlas (TCGA) database (n=335) and Gene Expression Omnibus (GEO) database (n=865). Unsupervised consensus clustering was used to identify subtypes of gastric adenocarcinoma. We performed functional enrichment analysis, immune infiltration analysis, drug sensitivity analysis, and molecular feature analysis to determine the clinical application for different subtypes. The univariate Cox regression analysis and the LASSO regression analysis were subsequently used to identify prognosis-related methylation regulators and construct a risk model. Through unsupervised consensus clustering, patients were divided into two subtypes (cluster A and cluster B) with different clinical outcomes. Cluster B included patients with a better prognosis outcome and who were more likely to respond to immunotherapy. We then successfully built a predictive model and found five methylation-related genes (CHAF1A, CPNE8, PHLDA3, SPARC, and EHF) potentially significant to the prognosis of patients. The 1-, 3-, and 5-year areas under the curve of the risk model were 0.712, 0.696, and 0.759, respectively. The risk score was an independent prognostic factor and had the highest concordance index among common clinical indicators. Meanwhile, the tumor microenvironment, sensitivity of chemotherapeutic drugs, molecular features, and oncogenic dedifferentiation differed significantly across the risk groups and subtypes. We classified patients with gastric adenocarcinoma based on methylation regulators, which has positive implications for first-line clinical treatment. The prognostic model could predict the prognosis of patients and help to promote the development of precision medicine.