Machine learning prediction of obesity-associated gut microbiota: identifying Bifidobacterium pseudocatenulatum as a potential therapeutic target.

Abstract

Background: The rising prevalence of obesity and related metabolic disorders highlights the urgent need for innovative research approaches. Utilizing machine learning (ML) algorithms to predict obesity-associated gut microbiota and validating their efficacy with specific bacterial strains could significantly enhance obesity management strategies.

Methods: We leveraged gut microbiome data from 1,563 healthy individuals and 2,043 overweight patients sourced from the GMrepo database. We assessed the anti-obesity effects of Bifidobacterium pseudocatenulatum through experimentation with Caenorhabditis elegans and C3H10T1/2 cells.

Results: Our analysis revealed a significant correlation between gut bacterial composition and body weight. The top 40 bacterial species were utilized to develop ML models, with XGBoost demonstrating the highest predictive accuracy. SHAP analysis indicated a negative association between the relative abundance of six bacterial species, including B. pseudocatenulatum, and body mass index (BMI). Furthermore, B. pseudocatenulatum was shown to reduce lipid accumulation in C. elegans and inhibit lipid differentiation in C3H10T1/2 cells.

Conclusion: Bifidobacterium pseudocatenulatum holds potential as a therapeutic agent for managing diet-induced obesity, underscoring its relevance in microbiome-based obesity research and intervention.