Prostate cancer (PCa) is the second most prevalent malignancy in men globally, contributing significantly to disease burden. Early diagnosis is critical to patient treatment and prognosis. DNA methylation is among the earliest molecular changes in tumors, making it a vital biomarker for early diagnosis. Advances in liquid biopsy technology have enabled non-invasive diagnostics, particularly through the methylation analysis of circulating tumor DNA (ctDNA), facilitating early PCa detection. However, the low abundance of methylation signals in biological fluids presents challenges, as traditional PCR methods exhibit limited sensitivity, while high-throughput sequencing is costly and not conducive to widespread clinical use. Thus, there is an urgent need for sensitive, specific and cost-effective methylation detection methods to enhance early and precise PCa diagnosis.

Machine learning techniques were utilized to identify specific methylation sites from extensive PCa and normal control cohorts. A novel methylation-specific RPA system was developed to amplify bisulfite-treated ctDNA, subsequently enhanced by the CRISPR-Cas system to increase site specificity and biosignal amplification. Accordingly, a “Highly-specific ctDNA Methylation Liquid Biopsy” platform was established, and blood and urine were collected from 61 PCa patients and 30 healthy controls clinical trials.

The newly developed platform demonstrated feasibility for detection under thermostatic conditions, achieving a discrimination index above 500 and a detection limit below 0.005%. A panel of seven genes—ALOX12, ANGPTL2, GSTP1, HAPLN3, HOXD3, TAC1, and ADCY4—was identified, with significant differential methylation confirmed in plasma samples (P<0.05). A refined diagnostic panel comprising GSTP1, TAC1, and ADCY4 exhibited 88% sensitivity and 87% specificity in testing 91 samples of blood and urine. Stratification analyses indicated the potential of this platform for early PCa diagnosis (Gleason ≤7, 4