Microbiome centered prediction and prevention of recurrent infections

Antibiotics are a double-edged sword: they help clear the current infection, yet can also select for resistant pathogens, making future infections harder to treat.

While treatment guidelines recognize this collateral damage, we currently lack strategies to predict how treatments affect future recurrence and resistance at the individual patient level.

This problem is of particular importance in Urinary Tract Infections (UTIs); affecting the majority of women over their lifetime, UTIs can chronically recur despite antimicrobial treatment.

Importantly, UTIs are often self-seeded by strains residing in the gut microbiome, suggesting that the gut microbiome may provide means to predict current and future infections and could possibly even be manipulated to minimize infections.

Here, we propose an interdisciplinary approach combining high-throughput phenotyping and genomics of same-patient gut-microbiome and UTI samples with machine-learning analysis of clinical records, towards a look-ahead treatment strategy for recurrent infections.

First, we will use whole-genome and meta-genome approaches to sensitively detect infecting strains within the patients microbiome and develop a gene-based model for the infectivity of strains and thereby for the likely infecting agent and resistance profile of infection.

Second, we will use long-read sequencing to map genetic linkage among resistances in each patients microbiome, enabling the development of a reinforcement machine-learning model to assign treatments that minimize both the risk of treatment failure and of future resistance.

Finally, quantifying in vivo and in vitro the impacts of antibiotic intake on microbiome composition, we will test the feasibility of prescribing antibiotics that manipulate the microbiome in favor of less infectious strains.

Together, this unique research-to-clinic data-rich approach will establish the basic foundations for a microbiome-based paradigm of look-ahead treatment strategies.