Early detection of urological tumors based on genomic characteristics of cell-free DNA fragments: a multi-center study.

Cell-free DNA (cfDNA) has shown potential in distinguishing cancer patients from healthy individuals. This study investigates cfDNA fragmentomics-fragmentation patterns, end motifs (EDMs), and breakpoint motifs (BPMs)-to develop an early detection method for bladder urothelial carcinoma (BLCA), prostate adenocarcinoma (PRAD), and clear cell renal cell carcinoma (ccRCC). Using low-coverage whole genome sequencing (lcWGS) on plasma samples from 758 participants (including BLCA, PRAD, ccRCC, benign prostatic hyperplasia patients, and healthy controls), we analyzed cfDNA features. Machine learning models (logistic regression, support vector machine, random forest, XGBoost, Stacking) distinguished urological tumors from non-tumor cases with AUCs of 96% (BLCA), 99% (ccRCC), 92% (PRAD), and 89% (pan-cancer). Key discriminators included 6-bp EDMs and BPMs. A proposed two-tier screening strategy combining pan-cancer and cancer-specific features offers a cost-effective, non-invasive approach for early detection with strong clinical potential.