Recipient independent high accuracy FMT prediction and optimization in mice and humans

Background: Some microbiome composition can be associated with negative outcomes, including among others, obesity, disease and the failure to respond to treatment. Microbiota manipulation or supplementation have been argued to restore the microbiome associated with a healthy condition. Fecal Microbiota Transplantation (FMT) is among the most popular microbiome intervention procedures. Current practices are to choose the transplanted microbiome based on the donor phenotype, and not based on the expected recipient phenotype. However, the two differ drastically. We here propose an algorithm to predict the expected outcome of FMT from the donor phenotype, and optimize the FMT for different required outcomes. Results We here show, using multiple microbiome properties, that the donor and recipient phenotypes differ widely, and propose a tool to predict the recipient phenotype after the FMT using only the donors' microbiome and when available demographics for transplants from humans to either antibiotic treated mice, or other humans. We then extend the method to optimize the best-planned transplant (bacterial cocktails) by combining the predictor and a genetic algorithm (GA). We validate the predictor using a de-novo FMT experiment highlighting the possibility to choose transplants that optimize an array of required goals. We further show that a limited number of taxa is enough to produce an optimal FMT. Conclusions Over the shelf FMT require recipient independent optimized FMT selection. Such a transplant can be from an optimal donor, or from a cultured set of microbes. We have here shown the feasibility of both types of donations in antibiotic treated mice and for transplants between humans.