Scholarlink

<p>Early detection of breast cancer (BC) is crucial for improving patient survival, yet current screening methods such as mammography exhibit limitations in sensitivity, especially in dense breasts. Circulating tumor DNA (ctDNA) methylation has emerged as a promising noninvasive biomarker for cancer detection. This study aimed to identify BC-specific DNA methylation markers from plasma ctDNA and develop a machine learning model for the early and accurate detection of BC. We constructed a BC-specific methylation profile by analyzing differentially methylated positions (DMPs) between 75 paired BC and adjacent normal tissues from TCGA. A total of 396 plasma samples were prospectively collected and divided into training (94 BC, 144 controls), validation (31 BC, 48 controls), and test sets (31 BC, 48 controls). A machine learning model was built using BC-specific DMPs, and its performance was evaluated via ROC analysis. Correlation between methylation levels and conventional biomarkers (CEA, ER) was assessed using Pearson’s coefficient. We identified 107 DMPs from TCGA that exhibited high variance (&gt; 0.001) in ctDNA. A model based on 21 DMPs achieved high accuracy in the training set (AUC = 0.972, sensitivity = 90.48%, specificity = 91.20%), with consistent performance in validation (AUC = 0.886) and test sets (AUC = 0.817). Methylation levels at four sites showed moderate to strong correlation with CEA and ER expression (all p &lt; 0.05). This study demonstrates that a ctDNA methylation-based machine learning model can effectively distinguish BC from non-cancerous conditions with robust sensitivity and specificity, supporting its potential as a complementary tool for early BC detection. These findings underscore the clinical relevance of ctDNA methylation signatures and highlight the promise of integrating liquid biopsy with machine learning in future cancer screening strategies. This study provides a novel methylation-driven framework for BC screening that complements existing biomarkers. Future work should validate these markers in larger multi-center cohorts and explore their utility in monitoring treatment response and recurrence.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Breast cancer detection via targeted enzymatic methyl sequencing of plasma cell-free DNA

  • 期刊论文

  • 发表日期：

    2026年4月1日

查看全文
  • Huiquan Su,
  • Yulu Liu,
  • Yihan Li,
  • Hailing Zheng,
  • Xingqiang Zhu,
  • Zhi he,
  • Hongyue Liao,
  • Yilong Lin,
  • Qingmo Yang,
  • Hongliang Chen

摘要

Early detection of breast cancer (BC) is crucial for improving patient survival, yet current screening methods such as mammography exhibit limitations in sensitivity, especially in dense breasts. Circulating tumor DNA (ctDNA) methylation has emerged as a promising noninvasive biomarker for cancer detection. This study aimed to identify BC-specific DNA methylation markers from plasma ctDNA and develop a machine learning model for the early and accurate detection of BC. We constructed a BC-specific methylation profile by analyzing differentially methylated positions (DMPs) between 75 paired BC and adjacent normal tissues from TCGA. A total of 396 plasma samples were prospectively collected and divided into training (94 BC, 144 controls), validation (31 BC, 48 controls), and test sets (31 BC, 48 controls). A machine learning model was built using BC-specific DMPs, and its performance was evaluated via ROC analysis. Correlation between methylation levels and conventional biomarkers (CEA, ER) was assessed using Pearson’s coefficient. We identified 107 DMPs from TCGA that exhibited high variance (> 0.001) in ctDNA. A model based on 21 DMPs achieved high accuracy in the training set (AUC = 0.972, sensitivity = 90.48%, specificity = 91.20%), with consistent performance in validation (AUC = 0.886) and test sets (AUC = 0.817). Methylation levels at four sites showed moderate to strong correlation with CEA and ER expression (all p < 0.05). This study demonstrates that a ctDNA methylation-based machine learning model can effectively distinguish BC from non-cancerous conditions with robust sensitivity and specificity, supporting its potential as a complementary tool for early BC detection. These findings underscore the clinical relevance of ctDNA methylation signatures and highlight the promise of integrating liquid biopsy with machine learning in future cancer screening strategies. This study provides a novel methylation-driven framework for BC screening that complements existing biomarkers. Future work should validate these markers in larger multi-center cohorts and explore their utility in monitoring treatment response and recurrence.