Scholarlink

<p>Esophageal cancer (ESCA) is a leading cause of cancer-related mortality in China, with over 95% of patients diagnosed at advanced stages, resulting in low survival rates. Early detection is critical for improving outcomes, yet current screening methods are invasive or lack sensitivity. We developed a non-invasive early-detection model by integrating circulating cell-free DNA (cfDNA) methylation and fragmentomics features. Using a three-layer stacked ensemble machine learning framework, we analyzed multicenter blood samples from 309 participants, including ESCA patients and high-risk individuals. The integrated model called ESim-seq (Esophageal Cancer Screening with Integrated Model) achieved an AUC of 0.994 (95% CI: 0.979–1.000) in validation. At 99% specificity, sensitivity reached 87.3%, with simulated screening showing a 24.8% increase in stage I detection. Stratified analyses confirmed robustness across demographics, risk factors, and tumor stages. Clinical benefit modeling projected a 56.3% stage I detection rate under ideal conditions, significantly improving survival outcomes compared to standard care. This study developed a cfDNA-based analytical framework and demonstrated the advantage of multi-feature integration for early detection of ESCA.</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

Non-invasive esophageal cancer screening in high-risk populations: a multi-omics integration model

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月23日

查看全文
  • Wenya Li,
  • Youbin Cui,
  • Jingcai Wu,
  • Yongzhi Wang,
  • Huidong Yu,
  • Xiaotong Xi,
  • Yu Ding,
  • Peipei Zhang,
  • Lei Ye,
  • Xiaoqiang Wang,
  • Xiaohuan Hou,
  • Zhenghua Liu,
  • Ling Lv,
  • Wanfu Men,
  • Xiangli Liu,
  • Yang Liu,
  • Mingrui Shao,
  • Xing Zhang,
  • Wanglong Deng,
  • Dongsheng Chen,
  • Yong Li,
  • Shun Xu

摘要

Esophageal cancer (ESCA) is a leading cause of cancer-related mortality in China, with over 95% of patients diagnosed at advanced stages, resulting in low survival rates. Early detection is critical for improving outcomes, yet current screening methods are invasive or lack sensitivity. We developed a non-invasive early-detection model by integrating circulating cell-free DNA (cfDNA) methylation and fragmentomics features. Using a three-layer stacked ensemble machine learning framework, we analyzed multicenter blood samples from 309 participants, including ESCA patients and high-risk individuals. The integrated model called ESim-seq (Esophageal Cancer Screening with Integrated Model) achieved an AUC of 0.994 (95% CI: 0.979–1.000) in validation. At 99% specificity, sensitivity reached 87.3%, with simulated screening showing a 24.8% increase in stage I detection. Stratified analyses confirmed robustness across demographics, risk factors, and tumor stages. Clinical benefit modeling projected a 56.3% stage I detection rate under ideal conditions, significantly improving survival outcomes compared to standard care. This study developed a cfDNA-based analytical framework and demonstrated the advantage of multi-feature integration for early detection of ESCA.