Distribution Characteristics of 10 Dominant Gut Microbiota in Patients with Liver Diseases Under Different Staging Systems

This study explores the potential of using the gut microbiota as a biomarker for liver disease classification by constructing machine learning classifiers. The classifiers were designed based on the abundance of 10 dominant bacterial taxa to categorize different liver disease scoring systems, including Child-Pugh score, model for end-stage liver disease (MELD), albumin-bilirubin (ALBI), fibrosis 4 score (FIB-4) and aspartate aminotransferase-to-platelet ratio index (APRI). Significant variations in gut microbiota composition were observed across various stages of liver disease. For example, the relative abundances of Enterococcus , Lactobacillus and Eubacterium rectale exhibited notable differences between cirrhotic patients with Child–Pugh grades A and B and grades A and C, and between patients with MELD scores of 6–15 and those with MELD scores of 15–40. In terms of the FIB-4 index, Enterococcus , Lactobacillus , Clostridium leptum , E. rectale and Faecalibacterium prausnitzii differed significantly across the low-, medium- and high-fibrosis groups. Analysis of the ALBI score revealed significant differences in the abundances of Enterococcus , Lactobacillus , Bacteroides , C. leptum , E. rectale and F. prausnitzii between grade 1 and grades 2–3. We constructed classifiers using machine learning algorithms based on the content of 10 dominant gut bacteria to classify the grading of different scoring systems. The highest area under the receiver operating characteristic (AUROC) values reported were for CP_XGBoost (0.7090 for Child–Pugh), MELD_SVM_UP (0.6346 for MELD), ALBI_XGBoost_SMOTE (0.7298 for ALBI), FIB4_SVM_UP (0.5873 for FIB-4) and APRI_SVM_UP (0.6826 for APRI). These findings highlight the potential of integrating gut microbiota analysis into existing liver disease scoring frameworks to increase diagnostic accuracy and improve patient care.