Discovery and validation of cell-free DNA methylation markers for specific diagnosis, differentiation from benign tumors, and prognosis of breast cancer

作者：Lijing Gao, Lei Zhang, Jinyin Liu, Liwan Wang, Qingzhen Fu, Siyu Liu, Yanbing Li, Chao Qu, Ning Zhao, Shiheng Tan, Fang Fang, Tian Tian, Shi Jiang, Jia Qian Wu, Yan Dong, Yajie Gong, Yukun Cao, Fan Wang, Xianyu Zhang, Da Pang, Yashuang Zhao · 发表于：Breast Cancer Research · 年份：2025 · DOI：10.1186/s13058-025-02066-x · 被引用次数：3 · 研究领域：Cancer Genomics and Diagnostics、Epigenetics and DNA Methylation、Cancer Cells and Metastasis

BACKGROUND: Plasma cell-free DNA (cfDNA) methylation is emerging as a non-invasive marker for various cancers. We aimed to identify specific methylation markers for diagnosis, differentiation from benign tumors, and prognosis of breast cancer (BC), which are essential for clinical decision-making yet seldom examined together. METHODS: BC-specific methylation markers were identified using an in-house 850K dataset combined with large-scale publicly available 450 or 850K datasets. Multiplex digital droplet PCR (mddPCR) assays were developed to detect methylation in cfDNA from 201 BC patients, 83 healthy donors, and 71 individuals harboring benign tumors. Diagnostic and prognostic performance were evaluated using logistic and Cox regression models, respectively. The basic mechanism of a selected gene was explored through in vitro experiments. RESULTS: We identified 21 BC-specific methylated CpG sites that distinguished BC from tumor-adjacent tissues with high diagnostic accuracy. In the cfDNA cohort, three mddPCR assays targeting eight markers achieved an area under the curve (AUC) of 0.856 (95% CI = 0.814-0.898) for distinguishing BC from healthy controls, and 0.742 (95% CI = 0.684-0.801) for differentiating BC from benign tumors. Notably, combining these methylation markers with mammography and ultrasound improved diagnostic performance, resulting in an AUC of 0.898 (95% CI = 0.858-0.938) for differentiating BC from benign tumors. In the TCGA-BC dataset, prognostic model based ...