Scholarlink

Background <p>Chronic liver disease (CLD) and its progression to compensated cirrhosis (CC), decompensated cirrhosis (DC), and acute-on-chronic liver failure (ACLF) represent a global health crisis with high mortality. The “gut-liver axis” plays a pivotal role in this progression, yet a systematic characterization of the dynamic microbial evolution across the entire disease spectrum remains elusive. We aimed to systematically characterize the gut microbiota’s dynamic evolution across the spectrum of chronic liver disease and to develop machine-learning models for predicting its critical transitions.</p> Methods <p>In this retrospective cohort study, 947 patients with liver disease (CLD, <i>n</i> = 464; CC, <i>n</i> = 159; DC, <i>n</i> = 260; ACLF, <i>n</i> = 64) were enrolled. The relative abundances of 10 dominant gut microbiota—including <i>Enterococcus</i>,<i> Lactobacillus</i>,<i> Clostridium leptum</i>,<i> Clostridium butyricum</i>,<i> Eubacterium rectale</i>,<i> Faecalibacterium prausnitzii</i>,<i> Bacteroides</i>,<i> Enterobacterium</i>,<i> Bifidobacterium</i>, and <i>Atopobium cluster</i>—were quantified via qPCR. Random Forest (RF) models were constructed to predict transitions between disease stages. The discriminatory efficacy of the <i>Enterococcus/Eubacterium rectale</i> Ratio (E/Er Ratio) was further evaluated using Receiver Operating Characteristic (ROC) analysis.</p> Results <p>Four distinct dynamic evolutionary patterns were identified. The pro-inflammatory/anti-inflammatory ratio increased 5.4-fold from CLD to ACLF. Random Forest models accurately predicted all disease transitions. The model for the DC-to-ACLF transition performed best (Area Under the Curve, AUC = 0.961), with clinical parameters (PT, TBil) being the strongest predictors. While the addition of microbial features yielded a modest incremental gain in AUC for the late-stage transition, indices such as the Specific-Butyrate-to-Total ratio were identified as key features, providing critical biological insights into the systemic failure of the gut-liver axis. The E/Er Ratio further served as a robust, non-invasive marker for identifying critical disease turning points.</p> Conclusion <p>Liver disease progression is characterized by a systematic shift in the gut microbiota from an anti-inflammatory, butyrate-rich state to a pro-inflammatory, pathogen-dominant environment. The integrated RF models and the E/Er Ratio provide a powerful, non-invasive framework for the early prediction and risk stratification of chronic liver disease progression, offering potential targets for gut-centered therapeutic interventions.</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

Dynamic gut microbiota shift in chronic liver disease progression: machine learning models and the E/Er ratio as a predictive biomarker

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月25日

查看全文
  • Dai Mengting,
  • Ji Zhaoyang,
  • Dong Jingjing,
  • Wu Jianbo,
  • Ye Tongtong,
  • Zhou Linling,
  • Qin Jiaying,
  • Luo Jialu,
  • Chen Ping,
  • Xu Mingzhi

摘要

Background

Chronic liver disease (CLD) and its progression to compensated cirrhosis (CC), decompensated cirrhosis (DC), and acute-on-chronic liver failure (ACLF) represent a global health crisis with high mortality. The “gut-liver axis” plays a pivotal role in this progression, yet a systematic characterization of the dynamic microbial evolution across the entire disease spectrum remains elusive. We aimed to systematically characterize the gut microbiota’s dynamic evolution across the spectrum of chronic liver disease and to develop machine-learning models for predicting its critical transitions.

Methods

In this retrospective cohort study, 947 patients with liver disease (CLD, n = 464; CC, n = 159; DC, n = 260; ACLF, n = 64) were enrolled. The relative abundances of 10 dominant gut microbiota—including Enterococcus, Lactobacillus, Clostridium leptum, Clostridium butyricum, Eubacterium rectale, Faecalibacterium prausnitzii, Bacteroides, Enterobacterium, Bifidobacterium, and Atopobium cluster—were quantified via qPCR. Random Forest (RF) models were constructed to predict transitions between disease stages. The discriminatory efficacy of the Enterococcus/Eubacterium rectale Ratio (E/Er Ratio) was further evaluated using Receiver Operating Characteristic (ROC) analysis.

Results

Four distinct dynamic evolutionary patterns were identified. The pro-inflammatory/anti-inflammatory ratio increased 5.4-fold from CLD to ACLF. Random Forest models accurately predicted all disease transitions. The model for the DC-to-ACLF transition performed best (Area Under the Curve, AUC = 0.961), with clinical parameters (PT, TBil) being the strongest predictors. While the addition of microbial features yielded a modest incremental gain in AUC for the late-stage transition, indices such as the Specific-Butyrate-to-Total ratio were identified as key features, providing critical biological insights into the systemic failure of the gut-liver axis. The E/Er Ratio further served as a robust, non-invasive marker for identifying critical disease turning points.

Conclusion

Liver disease progression is characterized by a systematic shift in the gut microbiota from an anti-inflammatory, butyrate-rich state to a pro-inflammatory, pathogen-dominant environment. The integrated RF models and the E/Er Ratio provide a powerful, non-invasive framework for the early prediction and risk stratification of chronic liver disease progression, offering potential targets for gut-centered therapeutic interventions.