跨15种癌症类型和7种亚型的甲基化特征比较

Differentially methylated regions (DMRs), the regions that have a different methylation pattern between tumor and non-tumor genomes, can serve as accurate biomarkers for cancer detection. Here we compare DMRs across 15 cancer types and 7 cancer subtypes. In this retrospective study, tissue samples from 832 patients with 15 different cancer types, including gastrointestinal (GI), genitourinary (GU), gynecological, breast, lung, melanoma, and head & neck cancers, were analyzed to measure methylation levels across genomic regions. Advanced adenoma (AA), an important precursor to colorectal cancer, was also included as a comparator among GI indications. Sets of DMRs were identified in each cancer type based on comparison with methylation levels of healthy samples. DMR signatures were compared across cancer types and subtypes by calculating pairwise overlaps. Across all 15 cancer types, the overlap of shared DMRs ranged from 13% (melanoma vs gastric cancer) to 61% (esophageal cancer vs gastric cancer). Notably, we found strong overlap among GI indications, a group including esophageal, gastric, small intestine, liver, pancreatic, and colorectal cancer, as well as AA. Compared to non-GI cancers, the pairwise overlap of DMRs in GI indications was, on average, 17% higher. Moreover, 65% of DMRs were shared between AA and colorectal cancer, which represented the highest overlap of shared DMRs in any pairwise comparison in this analysis. Among breast and lung cancer samples, we compared subtypes and found strong, but variable, overlap. For instance, among four major subtypes of breast cancer, we found overlaps ranging 53%-68% between subtypes. Among three lung cancer subtypes, overlaps ranged from 34% to 44%. In this study, we demonstrate that each cancer's DMR signature is ultimately unique and can be used as a biomarker for cancer detection even though there is some overlap between DMR signatures of different cancer types. A deeper study of the shared and unique DMR signatures may provide insights into underlying biological mechanisms of cancer development. For example, the high overlap of DMR signatures among GI cancers suggests some shared mechanisms; similarly, while DMR signatures are expected to be similar between subtypes of a cancer, investigation into the patterns unique to each subtype may help resolve observed differences in clinical prognosis and treatment response.