Detection of early stage pancreatic cancer using 5-hydroxymethylcytosine signatures in circulating cell free DNA

Pancreatic cancers are typically diagnosed at late stage where disease prognosis is poor as exemplified by a 5-year survival rate of 10%. Earlier diagnosis would be beneficial by enabling surgical resection or earlier application of therapeutic regimens. We investigated non-invasive detection of pancreatic ductal adenocarcinoma (PDAC) by interrogating changes in 5-hydroxymethylation cytosines (5hmC) of circulating cell free DNA in the plasma of a PDAC cohort (n=64) in comparison with a non-cancer cohort (n=243). 5hmC density is reduced in promoters and enhancers, whereas it is increased over 3’UTR, intron and TTS regions in PDAC compared to non-cancer cfDNA. Differential hydromethylation is rampant and found in thousands of genes. Stringent filtering by significance reveals genes related to pancreas function or development (GATA4, GATA6, PROX1, ONECUT1, MEIS2), and/or cancer pathogenesis (YAP1, TEAD1, PROX1, ONECUT1, ONECUT2 and IGF1). Furthermore, we observe enrichment for genes that are de-regulated upon activation of KRAS and inactivation of TP53, both of which are commonly observed in PDAC tumors. Regularized regression models, built using 5hmC densities in genes with the most variable 5hmC counts, performed with AUC of 0.92 on discovery dataset and 0.92-0.94 on two independent test sets. Furthermore, investigation of 5hmC occupancy in PDAC tumor tissue revealed that tissue-derived features can be used to accurately classify PDAC cfDNA (AUC=0.88). These findings suggest that 5hmC changes, at least partially derived from tumor tissue, enable classification of PDAC patients with high fidelity and are worth further investigation on larger cohorts of patient samples. Examination of the 5-hydroxymethyl cytosine (5-hmC) profiles from cell-free DNA from pancreatic adenocarcinoma in comparison to non-cancer patients. Inclusion of PDAC tumor samples and normal pancreas tissue 5-hmC profiles and ChIP-seq. **Submitter declares that raw data are not provided for this submission due to patient privacy concerns. Fastq files can be made available upon written request for submission to the study institutional review board for approval.**