Probiotics mitigate non-antibiotic drug-induced dysbiosis via protein homology-driven competitive binding.

While the impact of non-antibiotic drugs on gut bacteria is well-known, their mechanisms of action remain poorly characterized, and effective mitigation strategies for drug-induced dysbiosis are still limited. Here, we screened bacteria-derived drug-target protein homologs (BDTPHs) mapped to 63 target proteins and 107 associated drugs to quantify "drug-BDTPH-bacterium" interactions. These interactions were validated by co-culture experiments using 10 drugs and 25 strains, enzyme assays, and genetic perturbations in Escherichia coli. Ex vivo and in vivo testing with six drugs showed that over 50% of affected genera exhibited high affinity, indicating microbiota alterations through the "drug-BDTPH-bacterium" axis. Leveraging this quantitative interaction framework, we identified a strain of Bifidobacterium animalis that can competitively bind methotrexate through high-affinity BDTPH, thereby effectively alleviating gut microbiota dysbiosis in vivo. Our findings elucidate a mechanism by which non-antibiotic drug effects on bacterial growth, and suggest a universal homology-based competition strategy to restore drug-disrupted microbiota.