PubAg

Main content area

Ensemble optimization of microbial conversion of glycerol into 1, 3-propanediol by Klebsiella pneumoniae

Author:
Pan, Duo-Tao, Wang, Xu-Dong, Shi, Hong-Yan, Yuan, De-Cheng, Xiu, Zhi-Long
Source:
Journal of biotechnology 2019 v.301 pp. 68-78
ISSN:
0168-1656
Subject:
Klebsiella pneumoniae, computer simulation, fermentation, glycerol, kinetics, mathematical models, prediction, sampling, uncertainty
Abstract:
Using mathematical model and computer simulation to predict biological processes and optimize the target production is an important strategy for optimizing fermentation process. However, the inherent uncertainty of the kinetic model severely limits the predictive capability. In this study, optimize target production, such as productivity and yield of 1, 3-propanediol produced by Klebsiella pneumoniae using glycerol as substrate, the ensemble modeling approach was used to reduce the model’s uncertainty for fermentation process as much as possible, and effectively improve its prediction performance. Firstly, through sensitivity analysis, the parameters having significant influence on the model were determined as the adjustable parameters for the ensemble modeling. After comparison, the appropriate threshold coefficient of the model error was determined, and the sampling method was used to generate as many equivalent parameter sets as possible. Each set of parameters was separately applied for the simulation, and all the predicted values were integrated for the weighted average. Therefore, the expected value of the prediction was obtained. Compared with the traditional simulation using single parameter set, the ensemble modeling method achieved the lower relative error between the prediction and the experimental value and the greatly improved model prediction performance. Moreover, the optimal productivity and yield of 1, 3-propanediol and the corresponding operating conditions were obtained, respectively. The ensemble modeling approach effectively compensates for the uncertainties of the model, making its prediction performance more practical, which is important for computer simulations to predict and guide the actual production process.
Agid:
6466255