Colorectal cancer (CRC) is a major cause of cancer-related deaths. Screening colonoscopy is the standard of care for CRC detection, but compliance remains below 50% because of challenging preps, procedural costs and potential complications. Specific mutations are rare in CRC, precluding their utility as screening biomarkers. Epigenetic changes, including DNA 5-methylcytosine (5mC), contribute to CRC and are interrogated in several FDA-approved tests, but have limited sensitivity and no prognostic information. Sensitive blood tests for biomarkers in circulating cell-free DNA (cfDNA) could offer greater convenience and higher compliance. In addition to 5mC, changes in 5-hydroxymethylcytosine (5hmC) are important in normal and disease states. 5hmC is an abundant, stable modified cytosine that is generated by DNA demethylation and frequently marks active gene transcription. 5hmC and 5mC could serve as superior biomarkers, given their widespread genomic prevalence, distinct roles in gene regulation, and robust chemical stability. We developed a highly sensitive and selective chemical labeling technology (nano-hmC-Seal) to capture 5hmC bases, and using next generation sequencing (NGS), map their genomic positions. In preliminary studies, we captured and profiled 5hmC and 5mC in cfDNA and paired tumor tissues from 47 controls and 136 patients with recently diagnosed cancers, including 46 with CRC. We developed a classifier to identify genes with differential 5hmC changes in CRC patients and identified consistent cancer-specific epigenetic signatures in tumors and cfDNA. Because of the high sensitivity (~1ng DNA), robustness, clinical convenience and cost-efficiency, we hypothesize that unique cell-free 5hmC and 5mC profiles might be useful as plasma biomarkers to detect CRC and ultimately predict tumor prognosis. We propose to use PLCO biorepository samples and our own archived CRC with annotated clinical outcomes, and newly recruited prospective CRC patients to refine and validate the sensitivity of differentially expressed 5hmC and 5mC gene signatures for CRC diagnosis and prognosis. We propose: Aim 1 to profile 5hmC and 5mC in cfDNA from PLCO samples (CRC and control), using nano-hmCSeal and NGS and refine algorithms for sensitive CRC blood test, using 200 CRC cases and 400 controls for training set, and an additional 200 cases and 400 controls from our prospective cohort for validation set. We will develop discriminators of 5hmC and 5mC modifications to diagnose CRC and control analyses by tumor location, stage, patient age and gender that likely modulate 5hmC and 5mC distributions; Aim 2a to profile 100 PLCO tumors for 5hmC and 5mC distributions and compare distributions to cfDNA (aim 1 samples); In aim 2b, we will study 150 tumors from archived samples with recurrence data (60 relapsed patients vs. 90 clinical feature-matched non-relapsed patients), to detect a predictive index for tumor prognosis. In aim 2c, we will compare markers determined in tumors to markers detected in cfDNA to assess whether we can develop a minimally-invasive, cost-efficient, clinically convenient cfDNA-based marker for tumor prognosis.