Scholarlink

Inflammatory bowel disease (IBD), comprising ulcerative colitis (UC) and Crohn’s disease (CD), presents significant diagnostic challenges. To address this, we developed and applied a multi-model machine learning (ML) pipeline integrating LASSO regression, Multi-layer Perceptron (MLP) neural networks, and Support Vector Machine Recursive Feature Elimination (SVM-RFE) to gene expression data sourced from the Gene Expression Omnibus (GEO) database. Our comprehensive computational workflow involved rigorous data preprocessing, identification of differentially expressed genes (DEGs), sophisticated feature selection, and the development of predictive classification models. This ML-driven analysis identified key diagnostic gene biomarkers: DPP10, LCN2, S100A8, and S100P. Predictive models based on DPP10 and LCN2 demonstrated exceptional diagnostic performance, achieving area under the curve (AUC) values exceeding 0.94. S100A8 and S100P also exhibited significant diagnostic potential. These results underscore the efficacy of integrated ML approaches in extracting robust diagnostic signatures from complex genomic data. The identified biomarkers and the computational strategy provide a foundation for developing more accurate, data-driven diagnostic tools for IBD.

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

Deciphering Diagnostic Biomarkers and Characteristics in Ulcerative Colitis and Inflammatory Bowel Disease Through Genetic Profiling

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Zimo Wang,
  • Tangyu Yuan,
  • Xinxia Song,
  • Xiaoning Li,
  • Xiaowei Han,
  • Xuemei Zheng,
  • Jiayi Liu,
  • Jing Ban,
  • Ziang Pang,
  • Lei Yang,
  • Pengtao Liu

摘要

Inflammatory bowel disease (IBD), comprising ulcerative colitis (UC) and Crohn’s disease (CD), presents significant diagnostic challenges. To address this, we developed and applied a multi-model machine learning (ML) pipeline integrating LASSO regression, Multi-layer Perceptron (MLP) neural networks, and Support Vector Machine Recursive Feature Elimination (SVM-RFE) to gene expression data sourced from the Gene Expression Omnibus (GEO) database. Our comprehensive computational workflow involved rigorous data preprocessing, identification of differentially expressed genes (DEGs), sophisticated feature selection, and the development of predictive classification models. This ML-driven analysis identified key diagnostic gene biomarkers: DPP10, LCN2, S100A8, and S100P. Predictive models based on DPP10 and LCN2 demonstrated exceptional diagnostic performance, achieving area under the curve (AUC) values exceeding 0.94. S100A8 and S100P also exhibited significant diagnostic potential. These results underscore the efficacy of integrated ML approaches in extracting robust diagnostic signatures from complex genomic data. The identified biomarkers and the computational strategy provide a foundation for developing more accurate, data-driven diagnostic tools for IBD.