Machine Learning–Assisted Metabolic Characteristics for Revealing the Key Functionality of Wholemeal Sourdough Bread

This study employed metabolomic profiling, in vitro cell models, and transcriptomic analysis to investigate the protective effects and mechanisms of metabolites from lactic acid bacteria (LAB)–fermented wholemeal sourdough bread against lipopolysaccharide (LPS)‐induced hepatocyte injury. Machine learning approaches, including graph neural networks (GNNs), were applied to decipher the complex metabolomic and transcriptomic data. Metabolomic analysis revealed that LAB fermentation significantly altered the metabolite profiles of wholemeal sourdough bread, leading to enriched levels of phenolic antioxidants, flavor compounds, and key intermediates of energy metabolism. These changes were particularly pronounced in bread fermented with Pediococcus pentosaceus J28 and Lactobacillus fermentum YC‐22. Building on these metabolic insights, this study further evaluated the hepatoprotective effects of extracts from these types of fermented bread on LPS‐induced injury in BRL‐3A hepatocytes. The extracts significantly attenuated hepatocyte damage, enhanced antioxidant enzyme activities, reduced reactive oxygen species (ROS), and improved mitochondrial membrane potential (MMP). Transcriptomic analysis revealed that these effects were mediated through the downregulation of calcium transport–related genes, leading to restored calcium homeostasis in the cytoplasm and endoplasmic reticulum and suppressing MAPK‐mediated inflammatory and apoptotic pathways. Integrative metabolomic–transcriptomic analysis further identified key metabolites (e.g., N‐acetyl‐L‐methionine and N‐acetyl‐DL‐tryptophan) that showed strong correlations with gene expression patterns, highlighting their potential roles in regulating redox and calcium homeostasis. These findings underscore the potential of LAB‐fermented wholemeal sourdough bread as a functional food for mitigating oxidative stress and liver inflammation.