5-hydroxymethylcytosine analysis reveals stable epigenomic changes in tumor tissue that enable cancer detection in cell-free DNA.

5-hydroxymethylcytosine (5hmC) is a stable epigenomic mark associated with gene regulation and tumorigenesis. To evaluate its potential for multi-cancer detection, epigenomic distribution of 5hmC was analyzed in tumor tissues (n = 217), normal tissues (n = 50), and cell-free DNA (cfDNA) from cancer (n = 1009) and non-cancer subjects (n = 1678) across breast, colon, lung, ovarian, and pancreatic cancers. Differential analysis revealed extensive redistribution of 5hmC in early-stage tumors that persisted into the late-stage, while global 5hmC abundance decreased in all tumors. Two distinct patterns of hydroxymethylation distribution were identified: one common across cancers and another tissue specific. Differential hydroxymethylation regions (DhmRs) that were tumor tissue specific had significantly improved cancer prediction in cfDNA. A multinomial logistic regression model was trained on tissue-specific DhmRs to predict Tissue of Tumor Origin (TOTO) of cfDNA with 85.2% accuracy. These findings demonstrate that stable, cancer-specific 5hmC signatures appear to be established early in tumorigenesis persist throughout disease progression, suggesting 5hmC as a potential pan-cancer marker for early and late-stage tumor detection via liquid biopsy.