Identification of gut microbial bile acid metabolic enzymes via an AI-assisted pipeline.

Summary

Using an AI-assisted workflow (BEAUT), the authors predicted over 600,000 gut microbial bile acid metabolic enzymes and released the HGBME database. They experimentally characterized new enzymes, including MABH and 3-acetoDCA synthetase (ADS); ADS produces a previously unreported skeleton bile acid (3-acetoDCA) via C–C bond extension, which is widespread and modulates gut microbial interactions.

Key Findings

• Developed the AI-assisted BEAUT workflow and compiled the HGBME database with >600,000 candidate microbial bile acid enzymes.
• Identified and validated previously uncharacterized enzymes, including monoacid acylated bile acid hydrolase (MABH) and 3-acetoDCA synthetase (ADS).
• Discovered a previously unreported skeleton bile acid (3-acetoDCA) produced by ADS via carbon–carbon bond extension; 3-acetoDCA is widespread and modulates gut microbial interactions.

Clinical Implications

While not immediately clinical, the resource enables targeting of specific microbial bile acid enzymes to modulate host bile acid pools, with potential applications in metabolic, cholestatic, and endocrine diseases.

Limitations

• Primarily preclinical mechanistic work without direct clinical outcomes.
• Most predicted enzymes remain to be experimentally validated and linked to human disease phenotypes.

Future Directions

Systematic validation of additional predicted enzymes, mapping enzyme–phenotype associations in human cohorts, and developing small-molecule or dietary strategies to modulate bile acid enzymatic activities.