Decoding Microbial Community Dynamics: Mechanistic Models, Machine Learning, and Active Learning
University of Wisconsin-Madison, Duke University
Envision a world where the trillions of bacteria inhabiting our bodies become the frontier of personalized medicine, having the ability to shape our health, performance and even influence behavior. When in harmony, this teeming world of bacteria offers numerous health benefits. However, a shift in this delicate balance can lead to substantial negative health effects due to contrasting evolutionary objectives. Precision engineering of the gut microbiome that can add, remove or modify functional capabilities of the system holds tremendous therapeutic potential for personalized and precision medicine. However, the complexity of this system that encompasses hundreds of species, unknown interaction networks and mechanisms driving these interactions have precluded our ability to effectively manipulate this system to our benefit. By integrating bottom-up construction of microbial communities with mechanistic, machine learning, hybrid models and active learning frameworks, we develop the capabilities to predict community dynamics and functions and optimize health-beneficial properties. Our work provides a foundation for exploring and exploiting the interaction networks driving microbial communities for precision medicine and beyond.