Urine-Based cfDNA Ensemble Modeling for Early Detection of Bladder Cancer Using Whole-Genome Methylation Sequencing.

Area of Science:

• Genomics
• Oncology
• Biomarker Discovery

Background:

• Early bladder cancer detection is challenging due to low tumor DNA in liquid biopsies.
• Biofluid selection and signal integration are critical for improving detection sensitivity.

Purpose of the Study:

• To develop and evaluate a urine-based bladder cancer detection framework using Enzymatic Methyl-seq (EM-seq).
• To assess the performance of integrating methylation and copy number variations (CNV) for bladder cancer diagnosis.

Main Methods:

• Analyzed EM-seq data from matched urine-plasma pairs to compare biofluid utility.
• Developed a urine-based detection model using XGBoost, combining methylation and CNV data.
• Validated the model on an independent test set of urine samples from bladder cancer patients and healthy controls.

Main Results:

• Urine samples showed higher tumor fractions and better concordance with tissue methylation than plasma.
• The urine-based model achieved 91.9% sensitivity and 80% specificity for bladder cancer detection (AUC=0.932).
• The model demonstrated high performance for non-muscle invasive bladder cancer (NMIBC) detection (AUC=0.928) and detected mutation-negative cases.

Conclusions:

• Urine-based EM-seq is a sensitive approach for early bladder cancer detection.
• Combining methylation and CNV data enhances diagnostic accuracy.
• This liquid biopsy method complements mutation-based detection strategies.