Abstract 4557: Genome-wide cfDNA hypo-methylation landscape of 500 patients, a real-world study

Abstract Background: Liquid biopsies analyzing cell-free DNA (cfDNA) from plasma and urine can offer clinical diagnostic information in a non-invasive way. One major application of DNA methylation in liquid biopsies focuses on targeted panel regions, aiming to detect hypermethylation in cancer related regions. Recent studies have revealed that the loss of DNA methylation is significantly enriched in late-replicating loci compared with early-replicating loci in cancer cells. In this study, we profiled genome-wide DNA methylation of plasma and urine samples from various of cancer indications to detect DNA methylation loss in tumor DNA. Methods: We used the PredicineEPIC assay for genome-wide DNA methylation profiling at 20× coverage. Unlike traditional differential methylation region (DMR) analysis, we identified fragment-level DNA methylation changes in CpG-rich genomic regions. For each fragment, we calculated the probability of the fragment being unmethylated versus methylated. This calculation is specific to genomic positions, with the background distribution built upon more than 100 healthy plasma samples. Results: Compared with genome-wide DNA methylation profile of healthy plasma samples, we observed two distinct types of methylation loss in cancer patient samples. In the first case, hypomethylation occurs at fragments with significant methylation loss, where the methylation status of all CpG sites transitions from methylated to unmethylated. We observed that the magnitude of methylation loss is positively correlated with tumor fraction. Based on wet-lab titration experiment results, we can achieve tumor fraction detection limit of 0.1% for prostate, breast, and lung cancers. In the second case, hypomethylation at the fragment level is more subtle, with only a portion of CpG sites becoming unmethylated. We found that many samples in this category had CHIP mutations or other methylation enzyme deficiencies detected, with DNMT3A, which encodes DNA methyltransferase, being the most frequent. The same method can also be applied to urine cfDNA data for the detection of bladder cancer. We analyzed over 500 whole-genome methylation samples across different cancer indications, including prostate, colon, pancreatic, lung, breast and bladder cancers. Overall, genome-wide DNA methylation loss is highly prevalent in cancer samples and varies across cancer types. Some chromosome regions are consistently hypo-methylated across samples and cancer indications. Conclusions: Our results indicate that global nuclear organization might be disrupted by methylation loss at closed chromatin compartments. The genome-wide methylation profile of circulating tumor DNA can measure cancer genome abnormality in terms of DNA methylation loss, CHIP mutations and copy number variations. Combination of all these features can provide a robust utility for cancer detection and estimating tumor fraction in liquid biopsies. Citation Format: Chao Dai, Giancarlo Bonora, Kemin Zhou, Shidong Jia, Pan Du. Genome-wide cfDNA hypo-methylation landscape of 500 patients, a real-world study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 4557.