Learning Microbial Metabolic Interactions and Their Evolution Through Artificial Intelligence

Microbes are an essential part of all ecosystems, influencing material flow and shaping their surroundings.
Understanding their mechanism of action is vital for addressing current challenges. Metagenomics studies provide insight into composition and function of microbiomes but predicting interactions within microbial communities and their stability and evolution from genotype remains challenging. Flux Balance Analysis and dynamic FBA have been successful in predicting microbial metabolic phenotypes. However, they rely on assumptions such as instantaneous biomass maximization in dFBA and are not always suitable for predicting long-term stable interactions. I will introduce this line of thinking, with focus on a recent novel algorithm that integrates artificial intelligence through reinforcement learning into metabolic models. The algorithm features a decision-making process that allows microbial agents to evolve by learning and adapting metabolic strategies for long-term fitness in the presence of other microbes. The method shows promise in elucidating microbial behavior, including phenomena like cooperative metabolite exchange in auxotrophs, their adaptation to rich environments, and effect of cheating in heterogenous communities.

Dr. Chan is an AP in the Dept. of Chemical and Biological Engineering at CSU. His research focuses on predicting microbiome metabolism using metabolic models. He works on the microbiome in the human gut, anaerobic digestion, and soil.