Cell-free DNA methylation profiles enable early detection of colorectal and gastric cancer

Colorectal cancer (CRC) and gastric cancer (GC) rank the top five common and lethal cancers worldwide. Early detection can significantly reduce the mortality of CRC and GC. However, current clinical screening methods including invasive endoscopic techniques and noninvasive fecal occult blood test screening tests/fecal immunochemical test have shown low sensitivity or unsatisfactory patient's compliance. Aberrant DNA methylation occurs frequently in tumorigenesis and cell-free DNA (cfDNA) methylation has shown the potential in multi-cancer detection. Herein, we aimed to explore the value of cfDNA methylation in the gastrointestinal cancer detection and develop a noninvasive method for CRC and GC detection. We applied targeted methylation sequencing on a total of 407 plasma samples from patients diagnosed with CRC, GC, and noncancerous gastrointestinal benign diseases (Non-Ca). By analyzing the methylation profiles of 34 CRC, 62 GC and 107 Non-Ca plasma samples in the training set (n=203), we identified 40,110 gastrointestinal cancer-specific markers and 63 tissue of origin (TOO) prediction markers. A new integrated model composed of gastrointestinal cancer detection and TOO prediction for three types of classification of CRC, GC and Non-Ca patients was further developed through logistic regression algorithm and validated in an independent validation set (n=103). The model achieved overall sensitivities of 83% and 81.3% at specificities of 81.5% and 80% for identifying gastrointestinal cancers in the test set and validation set, respectively. The detection sensitivities for GC and CRC were respectively 81.4% and 83.3% in the cohort of the test and validation sets. Among these true positive cancer samples, further TOO prediction showed accuracies of 95.8% and 95.8% for GC patients and accuracies of 86.7% and 93.3% for CRC patients, in test set and validation set, respectively. Collectively, we have identified novel cfDNA methylation biomarkers for CRC and GC detection and shown the promising potential of cfDNA as a noninvasive gastrointestinal cancer detection tool.