Abstract
Bladder cancer (BCa) is a significant health issue and poses a healthcare burden on patients, highlighting the importance of an effective detection method. Here, we developed a urine DNA methylation diagnostic panel for distinguishing between BCa and non-BCa. In the discovery stage, an analysis of the TCGA database was conducted to identify BCa-specific DNA hypermethylation markers. In the validation phase, DNA methylation levels of urine samples were measured with real-time quantitative methylation-specific PCR (qMSP). Comparative analysis of the methylation levels between BCa and non-BCa, along with the receiver operating characteristic (ROC) analyses with machine learning algorithms (logistic regression and decision tree methods) were conducted to develop practical diagnostic panels. The performance evaluation of the panel shows that the individual biomarkers of ZNF671, OTX1, and IRF8 achieved AUCs of 0.86, 0.82, and 0.81, respectively, while the combined yielded an AUC of 0.91. The diagnostic panel using the decision tree algorithm attained an accuracy, sensitivity, and specificity of 82.6%, 75.0%, and 90.9%, respectively. Our results show that the urine-based DNA methylation diagnostic panel provides a sensitive and specific method for detecting and stratifying BCa, showing promise as a standard test that could enhance the diagnosis and prognosis of BCa in clinical settings.