Artificial intelligence-driven donor-recipient gut microbiome matching for optimized fecal microbiota transplantation.

Post-FMT microbiome convergence of recipients toward donors is robustly associated with clinical efficacy across multiple diseases.

• Analysis was performed on multi-kingdom and functional profiles from pre- and post-FMT metagenomes
• Dataset included 515 FMTs across 30 cohorts and 12 diseases
• 94 metagenomes from 44 FMTs were newly collected for this study
• The association between convergence and clinical efficacy was observed across diseases, suggesting a generalizable mechanism

The MOZAIC framework achieved high predictive accuracy for post-FMT microbial convergence.

• MOZAIC stands for Microbiome Matching Optimization via Artificial Intelligence
• Average area under the curve (AUC) of 0.88 for forecasting microbiome convergence
• Accuracy and recall both exceeded 0.80 in forecasting microbiome convergence
• The framework integrates multi-dimensional donor-recipient microbiota features

MOZAIC achieved 78.7% accuracy in predicting clinical outcomes of FMT.

• Clinical outcome prediction was performed retrospectively
• The model was tested across multiple disease indications represented in the 30 cohorts
• Prediction of clinical outcomes was derived from the same multi-dimensional microbiota feature integration used for convergence prediction

Retrospective simulation using MOZAIC demonstrated a 1.44-fold improvement in clinical response rates over baseline matching.

• Baseline clinical response rate was 49.4%
• MOZAIC-optimized donor-recipient matching simulated a response rate of 71.0%
• This represents a 1.44-fold improvement over the unoptimized baseline
• The improvement was demonstrated through retrospective simulation rather than a prospective clinical trial

Multi-kingdom and functional microbiome profiles were analyzed to characterize donor-recipient microbiota features used in the matching framework.

• Analysis included multi-kingdom profiling (beyond just bacteria, potentially including fungi, viruses, etc.) and functional profiling
• Both pre- and post-FMT metagenomes were analyzed
• The multi-dimensional feature integration was central to MOZAIC's predictive performance
• Data spanned 30 cohorts and 12 distinct diseases

Microbiome convergence was established as a key mediator of FMT clinical efficacy.

• Convergence refers to the post-FMT shift of recipients' microbiome profiles toward those of their donors
• This mediating role was identified across a broad range of gastrointestinal and other diseases
• The finding supports microbiome convergence as a mechanistic explanation for variable FMT outcomes
• The authors describe this as establishing convergence as 'a key mediator of FMT'

MOZAIC is described as a scalable tool intended for precision matching in microbiota-based therapies beyond FMT.

• The framework is positioned as applicable to microbiota-based therapies broadly, not only FMT
• Scalability is highlighted as a key feature of the tool
• The study was based on retrospective data; prospective clinical validation is implied as a future step
• The framework integrates multi-dimensional donor-recipient microbiota features to enable precision matching