Explainable Artificial Intelligence to Decode the Blood-Brain Barrier Permeability of Gut Microbial Metabolites

Modulation of the nervous system by gut microbiota through metabolic pathways is a key mechanism of communication within the gut-brain axis. A critical factor determining whether gut microbial metabolites can exert functional effects in the brain is their ability to cross the blood-brain barrier (BBB). However, current methods for assessing BBB permeability lack systematic, standardized approaches and advanced predictive technologies. Traditional experimental techniques are often costly and time-consuming compared to computational methods. To address these limitations, we developed an automated molecular simulation workflow to generate a high-quality data set of gut microbial metabolites annotated with thermodynamic features related to BBB permeability. Based on this data set, we constructed an interpretable thermodynamic evaluation framework capable of accurately identifying key factors that influence transmembrane transport. The robustness and predictive power of our models were validated using two authoritative benchmark data sets, confirming their ability to reliably distinguish BBB-permeable from nonpermeable compounds. Furthermore, our findings highlight the substantial potential of gut microbiota metabolism to influence BBB permeability via metabolic pathways. Overall, this study provides a powerful tool for identifying gut microbiota-derived metabolites with potential biological activity in the brain and introduces a novel paradigm for the intelligent prediction of BBB permeability.