Accumulating evidence suggests that gut-microbiota metabolites contribute to human disease pathophysiology, yet the host receptors that sense these metabolites are largely unknown. Here, we developed a systems pharmacogenomics framework that integrates machine learning (ML), AlphaFold2-derived structural pharmacology, and multi-omics to identify disease-relevant metabolites derived from gut-microbiota with non-olfactory G-protein-coupled receptors (GPCRome). Specifically, we evaluated 1.68 million metabolite-protein pairs connecting 408 human GPCRs and 516 gut metabolites using an Extra Trees algorithm-improved structural pharmacology strategy. Using genetics-derived Mendelian randomization and multi-omics (including transcriptomic and proteomic) analyses, we identified likely causal GPCR targets (C3AR, FPR1, GALR1 and TAS2R60) in Alzheimer's disease (AD). Using three-dimensional structural fingerprint analysis of the metabolite-GPCR complexome, we identified over 60% of the allosteric pockets of orphan GPCR models for gut metabolites in the GPCRome, including AD-related orphan GPCRs (GPR27, GPR34, and GPR84). We additionally identified the potential targets (e.g., C3AR) of two AD-related metabolites (3-hydroxybutyric acid and Indole-3-pyruvic acid) and four metabolites from AD-related bacterium Eubacterium rectale, and also showed that tridecylic acid is a candidate ligand for orphan GPR84 in AD. In summary, this study presents a systems pharmacogenomics approach that serves to uncover the GPCR molecular targets of gut microbiota in AD and likely many other human diseases if broadly applied.