A Liquid Biopsy-Based Multi-methylation Marker Panel for Non-invasive Gastric Cancer Screening

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Background Gastric cancer (GC) is the fifth most prevalent cancer and the fourth leading cause of cancer-related mortality worldwide. The current gold standard for clinical diagnosis is gastroscopy, which, despite its high sensitivity and specificity, is limited by its invasive nature and high cost, making it unsuitable for large-scale screening. Furthermore, the diagnostic process lacks biomarkers that offer both high sensitivity and specificity. A screening model incorporating five methylation-based biomarkers (ELMO1, FGF12, NPY, SEPTIN9, ZNF671) was developed. Using these methylation profiles, GC risk prediction models were constructed employing Decision Tree, Logistic Regression, Random Forest, eXtreme Gradient Boosting (XGBoost), and support vector machine (SVM) algorithms. Results In the training cohort of 605 subjects (259 patients with gastric cancer, 346 controls), the model demonstrated an area under the curve (AUC) of 0.9585, accuracy of 87.93%, sensitivity of 81.85%, and specificity of 92.49%. In an independent validation cohort of 152 subjects (73 patients with gastric cancer, 79 controls), the model achieved an AUC of 0.8868, accuracy of 81.58%, sensitivity of 82.19%, and specificity of 81.01%. The model showed strong screening capability across various pathological stages (0 + IA + IB, IIA + IIB, IIIA + IIIB + IIIC, IV), with AUCs of 0.8210, 0.9149, 0.9357, and 0.9383, respectively. Validation results were consistent with those from the training cohort, indicating significant potential for early-stage detection. Conclusions This study establishes a non-invasive, peripheral blood DNA methylation-based detection method for GC screening. The model demonstrates robustness, high sensitivity, and specificity, offering an effective strategy for population-level screening.

Article activity feed