Scholarlink

<p>Metabolomics studies of heterogeneous diseases, such as hepatitis B virus (HBV) infection, liver cirrhosis (LC), and hepatocellular carcinoma (HCC), often grapple with small sample sizes that amplify biological variability and outlier biases, eroding statistical reliability. Conventional chemometric approaches, like orthogonal projections to latent structures-discriminant analysis (OPLS-DA), are susceptible to overfitting and spectral anomalies that hinder the detection of biomarkers. Here, we present a density-based resampling (DBR) framework designed to enhance the robustness of proton nuclear magnetic resonance (<sup>1</sup>H NMR)-based metabolomics analysis. DBR adaptively weights resampling probabilities according to local density in spectral space, prioritizing representative patterns over outliers. Applied to a cohort of 170 participants, DBR uncovered 12, 21, and 13 signature metabolites for HBV, LC, and HCC, respectively. Notable changes (<i>P</i> &lt; 0.05) spanned urea, malonate, choline, succinate, tyrosine, formate, and pantothenate across disease stages. Pathway enrichment pinpointed disruptions in phenylalanine, tyrosine, and tryptophan biosynthesis as a recurring theme. Moreover, DBR-facilitated clustering delineated distinct metabolic subtypes within each disease, illuminating underlying heterogeneity. This method eliminates the interference of outliers in biomarker searches and provides a robust toolkit for metabolomics in variable cohorts, paving the way for diagnostic markers and treatment options.</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

Density-Based Resampling Strategy Unveils Metabolic Heterogeneity and Subtypes in Liver Disease Progression

  • 期刊论文

  • 发表日期：

    2026年4月17日

查看全文
  • Yongpei Wang,
  • Xingxing Liu ,
  • Shuhai Lin ,
  • Kian-Kai Cheng ,
  • Lingli Deng,
  • Zhigang Luo,
  • Jiyang Dong

摘要

Metabolomics studies of heterogeneous diseases, such as hepatitis B virus (HBV) infection, liver cirrhosis (LC), and hepatocellular carcinoma (HCC), often grapple with small sample sizes that amplify biological variability and outlier biases, eroding statistical reliability. Conventional chemometric approaches, like orthogonal projections to latent structures-discriminant analysis (OPLS-DA), are susceptible to overfitting and spectral anomalies that hinder the detection of biomarkers. Here, we present a density-based resampling (DBR) framework designed to enhance the robustness of proton nuclear magnetic resonance (1H NMR)-based metabolomics analysis. DBR adaptively weights resampling probabilities according to local density in spectral space, prioritizing representative patterns over outliers. Applied to a cohort of 170 participants, DBR uncovered 12, 21, and 13 signature metabolites for HBV, LC, and HCC, respectively. Notable changes (P < 0.05) spanned urea, malonate, choline, succinate, tyrosine, formate, and pantothenate across disease stages. Pathway enrichment pinpointed disruptions in phenylalanine, tyrosine, and tryptophan biosynthesis as a recurring theme. Moreover, DBR-facilitated clustering delineated distinct metabolic subtypes within each disease, illuminating underlying heterogeneity. This method eliminates the interference of outliers in biomarker searches and provides a robust toolkit for metabolomics in variable cohorts, paving the way for diagnostic markers and treatment options.