Gut microbiota-based machine-learning signature for the diagnosis of alcoholic and nonalcoholic liver disease

Abstract

Abstract Background Machine learning (ML) algorithms improve diagnostic performance in various diseases. Alcoholic liver disease (ALD) and nonalcoholic fatty liver disease (NAFLD) account for majority of liver disease. Using gut microbiota-based ML algorithms, we evaluated the diagnostic index for ALD and NAFLD. Methods Fecal 16S rRNA sequencing data of 263 ALD (control, hepatitis, cirrhosis, and hepatocellular carcinoma [HCC]) and 201 NAFLD (control and hepatitis) subjects were collected. For the external validation, 126 ALD and 84 NAFLD subjects were recruited. Four supervised ML algorithms (support vector machine, random forest, multilevel perceptron, and convolutional neural network) were used for classification with 20, 40, 60, and 80 features, in which three nonsupervised ML algorithms (independent component analysis, principal component analysis, linear discriminant analysis, and random projection) were used for feature reduction. Results A total of 52 combinations of ML algorithms for each pair of subgroups were performed with 60 hyperparameter variations and 10-fold cross validation. ML models of convolutional neural network combined with principal component analysis achieved > 0.90 in the areas under the receiver operating characteristic curve (AUC). In the ALD, the diagnostic AUC values of the ML strategy (vs. control) were 0.94, 0.97, and 0.96 for hepatitis, cirrhosis, and HCC, respectively. The AUC values (vs. control) for NAFLD (hepatitis) was 0.93. In the external validation, AUC values of ALD and NAFLD (vs control) were > 0.90 and 0.88, respectively. Conclusion The gut microbiota-based ML strategy can be used for the diagnosis of ALD and NAFLD