abstract
- Human health is related to the gut microbiota, a complex ecosystem integrated by microorganisms and host cells. In vivo study of microbiota implies sophisticated analytical techniques, making it hard to study its dynamic changes. In vitro studies have proved helpful in deciphering interactions using representative synthetic microbial communities. Its characterization still depends on multiple sensors and analytical and molecular techniques. This work presents a SynCom including three bacteria with inulin as a fiber source to synthesize acetic, propionic, and butyric acids. A software sensor based on the Differential Neural Networks was implemented to estimate the SynCom dynamic. Species growth and byproduct concentration were estimated from available information (pH and optical density). We assessed the approach based on reported datasets; less than 5% of the error was achieved in less than 0.5 h, and we proved that our algorithm is suitable in noisy environments. This approach could accelerate the dynamic description of gut microbiota and reduce cost and time investment. © 2024 IEEE.