Turning the good against the bad: Machine learning to maximize disease-suppressing microbial interactions by optimizing nutrient conditions

• Applicant: Or Shalev Skriptchak, Ph.D.
• Subject Area: Medical Microbiology and Mycology, Hygiene, Molecular Infection Biology ; Bioinformatics and Theoretical Biology ; Microbial Ecology and Applied Microbiology
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 516931136

Antibiotics have served humanity with great success for almost a century, mitigating numerous infectious diseases. However, this outstanding success story is currently threatened because more and more pathogens evolve resistances against antibiotics. This makes formerly easy-to-treat diseases massive health threats again. Some researchers predict up to 10 million lethalities in 2050 caused by antibiotic resistance - a truly horrifying scenario. Given this unfolding health crises, we need novel ways to treat infections. The central goal of the proposed work is to develop such an alternative infection therapy that uses the power of microbial interactions to fight pathogens. Pathogens that attempt to colonize our body have to interact not only with our immune system, but also with the microbes that naturally live with us (commensals). Accordingly, commensals can protect us from pathogens by inhibiting their growth. I want to harness this natural protection mechanism by manipulating microbial interactions to turn native commensals specifically against pathogens. I will use a combination of high-throughput assays and machine learning to predict nutrients that cause the commensals to maximally inhibit pathogens. For that purpose I will develop an assay to measure >10,000 interactions per day between a diverse set of commensals and a given pathogen. This assay will provide me with big-data for training a machine learning model. I aim to use the trained model to predict whether unseen commensal bacteria can suppress pathogens based only on their genome and a given nutritional environment. Subsequently, I will use this trained machine learning model to obtain nutrient conditions that cause maximal suppression of the pathogen. Finally, I will test if the protection-inducing nutrient conditions can indeed protect a host (Caenorhabditis elegans) from a pathogenic infection. My goal is to develop a platform technology that allows finding nutrients to fight infections, and that can be easily modified and applied to basically any infection in the presence of commensals.This would demonstrate a minimal invasive method that would not introduce foreign microbes into the body. Moreover, no drugs with potential side effects have to be administered. Finally, this method could be individualized for every patient.

• DFG Programme: WBP Position