Building Predictive Understanding of Microbial Ecology by Bridging Microbial Growth Kinetics and Microbial Population Dynamics

Modeling microbial communities can provide predictive insights into microbial ecology, but current modeling approaches suffer from inherent limitations. In this study, a novel modeling approach was proposed to address those limitations based on the intrinsic connection between the growth kinetics of guilds and the dynamics of individual microbial populations. To implement the modeling approach, 466 samples from four full-scale activated sludge systems were retrieved from the literature. The raw samples were processed using a data transformation method that not only increased the dataset size by three times but also enabled quantification of population dynamics. Most of the 42 family-level core populations showed overall dynamics close to zero within the sampling period, explaining their resilience to environmental perturbation. Bayesian networks built with environmental factors, perturbation, historical abundance, population dynamics, and mechanistically derived microbial kinetic parameters classified the core populations into heterotrophic and autotrophic guilds. Topological data analysis was applied to identify keystone populations and their time-dependent interactions with other populations. The data-driven inferences were validated directly using the Microbial Database for Activated Sludge (MiDAS) and indirectly by predicting population abundance and community structure using artificial neural networks. The Bray-Curtis similarity between predicted and observed communities was significantly higher with microbial kinetic parameters than without parameters (0.70 vs. 0.66), demonstrating the accuracy of the modeling approach. Implemented based on engineered systems, this modeling approach can be generalized to natural systems to gain predictive understandings of microbial ecology.