Construction of a parsimonious kinetic model to capture microbial dynamics via parameter estimation

Abstract

Understanding microbial kinetic behaviour is important for bioprocessing engineering, such as chemical synthesis and bioremediation. However, development of proper models to capture complicated microbial kinetics is a challenging task. In this study, we demonstrate a comparative analysis of the complexity and accuracy tradeoff for modelling the growth of Shewanella oneidensis MR-1 in a batch culture. Based on a series of analyses, including residual analysis, scaled sensitivity coefficient analysis, parameter correlation analysis and the F-test, we estimated model parameters to construct a parsimonious Monod-based model that used the fewest parameters for best simulation of Shewanella growth using different carbon substrates. Sequential analysis was also applied to identify the time window for estimating each parameter in the kinetic model. This study shows that statistics-based parameter estimation is an efficient method to successively reconstruct and fine tune kinetic models for complex biological systems. IntroductionThe relation between microbial growth rates and substrate concentrations (often referred to as microbial growth kinetics) has been studied and substantiated for over a century.[1] Numerous models, including the Monod equation, [2] have been established to describe the growth kinetics. In spite of the wide applications of the Monod model, it does not always lead to a good description of a given bioprocess. One criticism of the Monod model is that the two parameters in the model, that is, the maximum specific growth rate (μ max ) and substrate saturation constant (K s ), can be correlated in estimation, which results in reported values that differ by several orders of magnitude.[1] In addition, the complicated culture conditions (e.g. growth with mixed substrates rather than a single substrate) make the original Monod model insufficient to reflect every aspect of growth kinetics. To overcome this problem, the model has to be refined by introducing

Identification of Dynamic Metabolic Flux Balance Models Based on Parametric Sensitivity Analysis

“…The least-squares methodology has been widely applied to estimate the parameter sets of many mathematical models used in engineering and for identifying biological models in particular . However, this method is only accurate if certain statistical assumptions are made. , On the other hand, the minimal sum of square errors is often considered a good indicator of model accuracy.…”