Scholarlink

<p>Early detection of pancreatic ductal adenocarcinoma (PDA) remains a major clinical challenge due to the lack of reliable biomarkers. We developed and validated a machine learning (ML)-based serum protein biomarker panel to enhance PDA diagnosis. Serum concentrations of 47 protein biomarkers were measured in 355 individuals using a Luminex™ bead-based immunoassay. Multiple ML algorithms were employed to construct a diagnostic model, with SHapley Additive exPlanations (SHAP) analysis used to determine the importance of each biomarker. The diagnostic performance of the panel was assessed by the area under the receiver operating characteristic curve (AUROC), F1 score, sensitivity, specificity, and accuracy, and further validated in an independent cohort of 130 individuals. Among the tested models, CatBoost demonstrated the highest diagnostic accuracy. SHAP analysis identified CA19-9, GDF15, and suPAR as key biomarkers, and the combined panel significantly outperformed CA19-9 alone in detecting PDA across all stages (AUROC 0.992 vs. 0.952) and in early-stage PDA (AUROC 0.976 vs. 0.868). Validation in another cohort confirmed the robustness of the model, with AUROC values of 0.977 for all stages and 0.987 for early-stage PDA. These findings suggest that ML-integrated biomarker panels may enable earlier and more accurate PDA detection in clinical practice.</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

Development of a serum protein biomarker panel for the diagnosis of pancreatic ductal adenocarcinoma using a machine learning approach

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月13日

查看全文
  • Dong Woo Shin,
  • Je-Yoel Cho,
  • Sukki Cho,
  • Yuna Youn,
  • Jin-Hyeok Hwang

摘要

Early detection of pancreatic ductal adenocarcinoma (PDA) remains a major clinical challenge due to the lack of reliable biomarkers. We developed and validated a machine learning (ML)-based serum protein biomarker panel to enhance PDA diagnosis. Serum concentrations of 47 protein biomarkers were measured in 355 individuals using a Luminex™ bead-based immunoassay. Multiple ML algorithms were employed to construct a diagnostic model, with SHapley Additive exPlanations (SHAP) analysis used to determine the importance of each biomarker. The diagnostic performance of the panel was assessed by the area under the receiver operating characteristic curve (AUROC), F1 score, sensitivity, specificity, and accuracy, and further validated in an independent cohort of 130 individuals. Among the tested models, CatBoost demonstrated the highest diagnostic accuracy. SHAP analysis identified CA19-9, GDF15, and suPAR as key biomarkers, and the combined panel significantly outperformed CA19-9 alone in detecting PDA across all stages (AUROC 0.992 vs. 0.952) and in early-stage PDA (AUROC 0.976 vs. 0.868). Validation in another cohort confirmed the robustness of the model, with AUROC values of 0.977 for all stages and 0.987 for early-stage PDA. These findings suggest that ML-integrated biomarker panels may enable earlier and more accurate PDA detection in clinical practice.