DECIPHER-PRAD: an advanced fragmentomics-based cell-free DNA assay for prostate cancer early detection

Early detection of prostate cancer is limited by the poor specificity of prostate-specific antigen (PSA)-based screening. Cell-free DNA (cfDNA) fragmentomics offers a promising non-invasive approach to improve screening accuracy and risk stratification. In this study, we enrolled 106 prostate cancer patients and 114 high-risk non-cancer individuals to develop a cfDNA fragmentomics-based screening assay using plasma whole-genome sequencing. Two fragmentomic features-copy number variation and fragment size profile-were incorporated into machine learning models for training and evaluated in an independent validation cohort of 83 cancer patients and 76 non-cancer individuals. The fragmentomics-based model achieved an area under the curve (AUC) of 0.933 in the training cohort (66.0% sensitivity at 95.6% specificity; 51.9% sensitivity at 98.2% specificity) with good calibration (slope: 0.957; intercept: 0.001), and maintained strong performance in the validation cohort (AUC: 0.887; 57.8% sensitivity at 92.1% specificity), showing rising predictive probabilities and sensitivity across advancing stages (Stage I-IV: 27.3% to 77.8%). Importantly, the model performed well in the PSA grey zone (4-10 ng/mL) with an AUC of 0.865 (69.0% sensitivity at 81.8% specificity). When integrated with total PSA levels, the combined algorithm achieved an AUC of 0.915 in the validation cohort and improved sensitivity at 98% specificity (Stage I-IV: 30.0% to 87.5%). These findings support the clinical potential of our cfDNA fragmentomic assay, particularly when combined with PSA, as a highly accurate and non-invasive tool for early prostate cancer detection.