5mC and 5hmC DNA alterations as sensitive and specific biomarkers for the non-invasive early detection of pancreatic ductal adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive malignancy and is rapidly becoming the 2nd leading cause of cancer-related deaths. The dismal 5-year survival rate of ~10% (the lowest of any major cancer) essentially stems from the fact that PDAC is diagnosed at late stages. Achieving early detection is crucial, offering a pivotal shift in staging that holds the promise of improved survival rates. Previous attempts to develop PDAC early detection tests have suffered from flaws in their study design, including 1) failure to enroll early-stage PDAC cases, 2) assuming that a biomarker developed in metastatic settings would also detect early-stage PDAC, or 3) assuming that a tissue-derived biomarker would also function in the blood. Therefore, we will develop and validate a bloodborne test in a cohort of patients enriched for early-stage PDAC. We will use the PLCO population to evaluate the lead time to PDAC diagnosis on pre-diagnosis biospecimens. Our test will be minimally invasive, highly sensitive, and cost-effective by harnessing the potential of 5methylcytosine (5mC) and 5-hydroxy methylcytosine (5hmC) modifications on circulating cell-free DNA. Importantly, commercially available tests based on 5mC demonstrated a high sensitivity for PDAC while a 5hmCbased test demonstrated specificity (but suboptimal sensitivity). Because 5hmC modifications occur genomewide and are abundant, they are biologically sensitive biomarkers, while 5hmC modifications are typically genetargeted and display >90% cancer- and stage-specificity. Therefore, a combination of the two approaches, one contributing sensitivity (5mC) and the other contributing specificity (5hmC) is desirable, innovative, and necessary to optimize the early detection of PDAC. Our goal is to develop a robust assay for detecting earlystage PDAC. Our design strategy will involve three aims. First, we will discover aberrantly methylated loci potentially diagnostic for PDAC with 5mC- and 5hmC-profiling in circulating cfDNA and matching tissues. Second, we will develop 5mC and 5hmC-based diagnostic biomarker signatures that discriminate patients with PDAC from controls using multiple advanced machine learning and statistical approaches. Here, we will train two machine learning algorithms—one for 5mC and another for 5hmC—to identify patients with PDAC based on qPCR results and then merge these two algorithms to create a detection signature for PDAC. Then, we will validate the 5mC, 5hmC, and 5mC+hmC qPCR assays in a large, cross-sectional cohort of biobanked samples from patients with PDAC vs. controls. Third, we will perform the 5mC, 5hmC, and 5mC+5hmC qPCR assays in prospectively collected samples from the PLCO study to determine how our assays may provide a positive PDAC signal much earlier than clinical diagnosis. In summary, this project aims to develop a highly sensitive, costeffective liquid biopsy for early detection of PDAC. If successful, this proposal will involve several innovations and will have the potential to transform clinical practice.