ABSTRACT: Owing to high cancer-specificity, DNA methylation alterations have emerged as front-runners in cell-free DNA (cf-DNA) biomarker development. However, much effort to date has focused on single cancers and have not explored the possibility of developing a pan-cancer diagnostic assay. Here, we undertook a genome-wide DNA methylation analysis for multiple gastrointestinal (GI) cancers, to develop a panGI diagnostic assay. By analyzing the DNA methylation data from 1940 tumor and adjacent normal tissues from TCGA and GSE72872 datasets, we first identified the differentially methylated regions (DMRs) between individual GI cancers and adjacent normal tissues, as well as across all GI cancers. We next prioritized a list of 67,832 tissue DMRs encompassing a 25.6 Mb genomic region by incorporating all significant DMRs across various GI cancers to design a custom SeqCap Epi, targeted bisulfite sequencing platform. Subsequent investigation of these tissue-specific DMRs in 300 cf-DNA specimens and applying machine learning algorithms led to the development of three distinct categories of DMR panels: 1) Cancer-specific biomarker panels with an AUC values of 0.98 (Colorectal cancer, CRC), 0.94 (Esophageal squamous cell carcinoma, ESCC), 0.90 (Esophageal adenocarcinoma, EAC), 0.90 (Gastric cancer, GC), 0.98 (Hepatocellular carcinoma, HCC), and 0.85 (Pancreatic ductal adenocarcinoma, PDAC); 2) A pan-GI panel that detected all GI cancers with an AUC of 0.90; and 3) A multi-cancer prediction panel, EpiPanGI Dx, with a prediction accuracy around 0.85-0.95 for most GI cancers. Utilizing a novel biomarker discovery approach, we provide first evidence for a cell-free DNA methylation biomarker assay that offer a robust diagnostic accuracy for all gastrointestinal cancers.
