The gut microbiome produces thousands of metabolites with potential to modulate central nervous system function through peripheral or direct neural mechanisms. Tourette syndrome, attention-deficit/hyperactivity disorder, and autism spectrum disorder exhibit shared neurotransmitter dysregulation and microbiome alterations, yet mechanistic links between microbial metabolites and receptor-mediated neuromodulation remain unclear. We screened 27,642 microbiome SMILES metabolites for blood–brain barrier permeability using rule-based SwissADME classification and a PyTorch 2.0 neural network trained on 7807 experimental compounds (test accuracy 86.2%, AUC 0.912). SwissADME identified 1696 BBB-crossing metabolites following Lipinski’s criteria, while PyTorch classified 2484 metabolites with expanded physicochemical diversity. Following 3D conformational optimization (from SMILES) and curation based on ≤32 rotatable bonds, molecular docking was performed against five neurotransmitter receptors representing ionotropic (GABRA2, GRIA2, GRIN2B) and metabotropic (DRD4, HTR1A) receptor classes. The top 50 ligands across five receptors demonstrated method-specific BBB classification (44% SwissADME-only, 44% PyTorch-only, 12% overlap), validating complementary prediction approaches. Fungal metabolites from Ascomycota dominated high-affinity top ligands (66%) and menaquinone MK-7 showed broad phylogenetic conservation (71.4% of phylum). Our results establish detailed receptor–metabolite interaction maps, with fungal metabolites dominating high-affinity ligands, challenging the prevailing bacterial focus of the microbiome and providing a foundation for precision medicine and a framework for developing microbiome-targeted therapeutics to address clinical needs in neurodevelopmental disorders.
Buendía-Corona et al. (Tue,) studied this question.