Fecal microbiota transplants are promising treatments for a wide range of gut and metabolic conditions, but results vary across patients. Now, researchers found that fecal microbiota transplant success may depend on how well a recipient’s gut microbiota shifts toward the donor’s microbiota after treatment.

The findings, published in Cell Reports, suggest that fecal microbiota transplants may work much better when donors are matched to recipients using AI tools that predict microbiota compatibility.

Fecal microbiota transplantation, or FMT, is an effective treatment for recurrent C. difficile infection and is being tested for many other conditions. Because treatment results vary widely between patients, past studies focused on finding “super donors,” but growing evidence suggests the recipients’ own microbiotas also affect success. 

Now, researchers led by Qi Su at the Chinese University of Hong Kong in China have analyzed microbiota data from 515 fecal microbiota transplants across 30 donor-recipient groups and 12 conditions.

Microbial shifts

Across hundreds of cases, FMT worked best when the microbiotas of recipients shifted to more closely resemble the donors’ microbiotas. 

Successful donor-patient pairs often started out more different from one another before treatment. Donor diversity alone did not predict success, while lower diversity in the recipients sometimes did, the team found. 

The way gut microbes changed after FMT varied depending on the condition being treated: some bacterial species increased in certain conditions but fell in others. FMT recipients who responded well tended to gain more donor-associated microbes, while non-responders often retained disrupted microbial communities. 

Clinical benefits

To better capture these patterns, the researchers created MOZAIC, an AI model designed to compare donor and recipients microbiotas in detail, including bacteria, fungi, viruses, and microbial functions. MOZAIC was trained and tested using separate datasets, and it predicted the microbiota convergence after FMT with an accuracy of about 80%.

The authors estimate that FMT response rates could rise from about 49% to 71% if donor selection were guided by the MOZAIC model. 

Because the study analyzed past data rather than testing the system in real patients, future trials should test MOZAIC directly and combine microbiota data with immune, metabolic, and genetic information, the authors say, “ultimately bridging the gap between microbial ecology and personalized medicine.”