Integrating a tailored recurrent neural network with Bayesian experimental design to optimize microbial community functions

Microbiomes interact dynamically with their environment to perform exploitable functions such as production of valuable metabolites and degradation of toxic metabolites for a wide range of applications in human health, agriculture, and environmental cleanup. Developing computational models to predict the key bacterial species and environmental factors to build and optimize such functions are crucial to accelerate microbial community engineering. However, there is an unknown web of interactions that determine the highly complex and dynamic behaviors of these systems, which precludes the development of models based on known mechanisms. By contrast, entirely data-driven machine learning models can produce physically unrealistic predictions and often require significant amounts of experimental data to learn system behavior. We develop a physically constrained recurrent neural network that preserves model flexibility but is constrained to produce physically consistent predictions and show that it outperforms existing machine learning methods in the prediction of experimentally measured species abundance and metabolite concentrations. Further, we present an experimental design algorithm to select a set of experimental conditions that simultaneously maximize the expected gain in information and target microbial community functions. Using a bioreactor case study, we demonstrate how the proposed framework can be used to efficiently navigate a large design space to identify optimal operating conditions. The proposed methodology offers a flexible machine learning approach specifically tailored to optimize microbiome target functions through the sequential design of informative experiments that seek to explore and exploit community functions. 1 Author summaryThe functions performed by microbiomes hold tremendous promise to address grand challenges facing society ranging from improving human health to promoting plant growth. To design their properties, flexible computational models that can predict the temporally changing behaviors of microbiomes in response to key environmental parameters are needed. When considering bottom-up design of microbiomes, the number of possible communities grows exponentially with the number of organisms and environmental factors, which makes it challenging to navigate the microbiome function landscape. To overcome these challenges, we present a physically constrained machine learning model for microbiomes and a Bayesian experimental design framework to efficiently navigate the space of possible communities and environmental factors.