Dietary Protein Source Shapes Gut Microbial Structure and Predicted Function: A Meta-Analysis with Machine Learning

Dietary proteins are major modulators of gut microbial ecology, yet the microbial signatures and functional consequences of plant-versus animal-based proteins remain poorly defined. Although digestibility and fermentation profiles differ by protein type, a systematic evaluation of how these differences shape microbial diversity, structure, and metabolic capacity is still lacking. This meta-analysis integrates machine learning and functional inference to identify microbial biomarkers and metabolic signatures associated with dietary protein source in murine models Following PRISMA guidelines, we analyzed 16S rRNA gene sequencing data from 10 murine studies (n = 187) comparing animal and plant protein diets. Alpha diversity was assessed using multiple indices, including Shannon, Inverse Simpson, and Chao1, to capture richness and evenness across samples. Beta diversity was evaluated using Aitchison distances. LEfSe and class-weighted Random Forest models were applied to identify differentially abundant taxa and predictive microbial features. Functional potential was inferred using PICRUSt2, and taxon–pathway relationships were explored through correlation and network analysis. Plant-protein diets significantly increased gut microbial diversity across all assessed alpha diversity metrics: Chao1 (p = 5.7 × 10 □¹□), Shannon (p = 3.3 × 10□□), and Inverse Simpson (p = 0.0013), reflecting enhanced species richness and evenness. These diets are associated with increased richness of 33 genera, including Bacteroides, Muribaculaceae, and Allobaculum, associated with SCFA production, nitrogen recycling, and bile acid metabolism. In contrast, animal-protein diets favored taxa such as Clostridium sensu stricto 1 and Colidextribacter, linked to proteolytic fermentation, ammonia, and sulfur metabolism. PCA of Aitchison distances revealed distinct microbial community structures between groups (ANOSIM R = 0.663, p < 0.001). Random Forest models achieved >88% accuracy (AUC = 0.995) in predicting dietary groups based on microbial composition. LEfSe confirmed key taxa distinguishing the dietary groups. Functional profiling revealed that plant-based diets enriched pathways for short-chain fatty acid (SCFA) production and aromatic amino acid degradation, while animal-based diets favored pathways involved in sulfur metabolism and branched-chain amino acid degradation. Network analysis identified Muribaculaceae as a central hub in plant-associated subnetworks linked to carbohydrate and SCFA metabolism, whereas Lactobacillus dominated animal-associated subnetworks enriched for proteolysis and sulfur-related functions. However, variation in dietary fat and fiber across studies may partially contribute to these microbial patterns, as these covariates were not consistently reported and therefore not included in the model. Dietary protein source significantly influences gut microbiota composition and function in murine models. This study identifies distinct taxonomic and functional signatures associated with plant and animal protein diets, with implications for microbial nitrogen utilization, sulfur metabolism, host–microbe interactions, and dietary modulation strategies. However, because other diet components, such as fat and fiber content, varied across studies and were not consistently reported, we could not include them as covariates. This represents a limitation in isolating protein-specific effects. Future studies should use controlled designs with matched non-protein components to address this gap