Machine learning-driven clinical decision support for liver cirrhosis: a gut microbiome-based web prediction model
Jianyuan Liu1,2,3,4, Shiran He1,3,4, Heng Zhang1,3,4
1School of Public Health, Guangxi Medical University, Nanning, Guangxi, China.
BMC Gastroenterology
|May 7, 2026
Summary
This study developed an interpretable machine learning model using gut microbes to predict liver cirrhosis (LC). The tool aids healthcare professionals in LC screening and early intervention.
Area of Science:
- Microbiome research
- Machine learning in medicine
- Bioinformatics
Background:
- Liver cirrhosis (LC) is a prevalent chronic liver disease with diagnostic limitations.
- Current LC diagnostic methods pose challenges in safety and accessibility.
Purpose of the Study:
- To develop an interpretable machine learning (ML) model for predicting liver cirrhosis (LC) using gut microbial data.
- To deploy the ML model as a web-based clinical decision support tool for LC screening.
Main Methods:
- Utilized bioinformatics and LEfSe analysis to identify key gut genera associated with LC.
- Employed LASSO regression for feature selection and Random Forest (RF) for model development.
- Validated the model using five-fold cross-validation and leave-one-dataset-out (LODO) analysis, with SHAP for interpretability and Streamlit for web deployment.
Main Results:
- Identified key gut genera (Veillonella, Lachnospira, Romboutsia, Akkermansia, Erysipelatoclostridium, Prevotella, UCG.005, Streptococcus) differentiating LC patients from controls.
- The Random Forest (RF) model achieved high predictive performance (AUC: 0.875 in cross-validation, 0.793 in LODO analysis).
- An online LC prediction tool was successfully deployed.
Conclusions:
- An interpretable RF model utilizing gut microbiome data shows potential for LC prediction.
- The web-based tool can assist healthcare professionals in LC screening and early clinical intervention.
