Ermira Lleshi et al.
Prostate cancer screening using prostate-specific antigen (PSA) has been shown to reduce mortality but with substantial overdiagnosis, leading to unnecessary biopsies. The identification of a highly specific biomarker using liquid biopsies, represents an unmet need in the diagnostic pathway for prostate cancer. In this study, we employed a method that enriches for methylated cell-free DNA fragments coupled with a machine learning algorithm which enabled the detection of metastatic and localised cancers with AUCs of 0.96 and 0.74, respectively. The model also detected 51.8% (14/27) of localised and 88.7% (79/89) of metastatic cancer patients in an external dataset. Furthermore, we show that the differentially methylated regions reflect epigenetic and transcriptomic changes at the tissue level. Notably, these regions are significantly enriched for biologically relevant pathways associated with the regulation of cellular proliferation and TGF-beta signalling. This demonstrates the potential of circulating tumour DNA methylation for prostate cancer detection and prognostication.