CN-121981320-A - Collaborative optimization system based on data-driven fusion metabolic network model and fermentation process mechanism analysis

CN121981320ACN 121981320 ACN121981320 ACN 121981320ACN-121981320-A

Abstract

The invention relates to the technical field of fermentation engineering, and particularly discloses a collaborative optimization system based on a data-driven fusion metabolic network model and fermentation process mechanism analysis, which comprises S1, process data acquisition and construction, S2, metabolic flux simulation and analysis, S3, multi-source data fusion, S4, machine learning modeling, S5, key metabolic node identification, S6, optimization strategy design and implementation, and systematic fusion and analysis of multi-source data in a fermentation process are realized by constructing a GEM-ML collaborative optimization framework, so that the limitation of a single method is overcome, key metabolic nodes influencing product synthesis can be identified more efficiently and accurately, and a targeted optimization strategy is generated according to the key metabolic nodes.

Inventors

TIAN XIWEI
XU FENG
CHU JU
ZHUANG YINGPING

Assignees

华东理工大学

Dates

Publication Date: 20260505
Application Date: 20260202

Claims (10)

1. A collaborative optimization system based on data-driven fusion metabolic network model and fermentation process mechanism analysis is characterized by comprising: the data acquisition and processing module is used for acquiring online process parameters and multi-source spectrum prediction data in the microbial fermentation process and constructing a fermentation process data set; the metabolism flux simulation module is used for carrying out flux balance analysis on the microbial metabolism network based on the genome scale metabolism model to obtain a metabolism flux data set; the data fusion module is connected with the data acquisition and processing module and the metabolic flux simulation module and is used for fusing the fermentation process data set and the metabolic flux data set to construct a mixed data set; The machine learning modeling and optimizing module is connected with the data fusion module and is used for constructing a target product yield prediction model by adopting a machine learning algorithm based on the mixed data set; the key node identification module is connected with the machine learning modeling and optimizing module and is used for identifying key metabolic reactions and metabolic pathways affecting target product synthesis from the prediction model by utilizing a feature importance analysis method; And the optimizing strategy output module is connected with the key metabolism node identification module and is used for generating and outputting a fermentation process optimizing strategy based on the key metabolism reaction and the metabolism path.
2. The collaborative optimization system based on the data-driven fusion metabolic network model and fermentation process mechanism analysis of claim 1, wherein the data fusion module matches metabolic flux data with process parameter data on a time scale by adopting a time alignment and interpolation method.
3. The collaborative optimization system based on the data-driven fusion metabolic network model and fermentation process mechanism analysis of claim 1, wherein in the machine learning modeling and optimization module, the machine learning algorithm is an Artificial Neural Network (ANN), the input is the mixed data set, and the output is the specific synthesis rate of a target product.
4. The collaborative optimization system based on the data-driven fusion metabolic network model and fermentation process mechanism analysis of claim 1, wherein the key node identification module outputs key metabolic reactions and SHAP values or weight scores thereof by adopting a SHAP analysis method.
5. The collaborative optimization system based on the data-driven fusion metabolic network model and fermentation process mechanism analysis according to claim 1, wherein the optimization strategy output module supports a Bayesian optimization or response surface method for determining an optimal feed proportioning.
6. The collaborative optimization system based on data-driven fusion metabolic network model and fermentation process mechanism analysis of claim 1, wherein the system is configured to optimize a fermentation process of Micromonospora echinocpora.
7. The collaborative optimization system based on the data-driven fusion metabolic network model and the fermentation process mechanism analysis according to claim 1 or 6 is characterized in that the target product is gentamicin C1a.
8. The collaborative optimization system based on the data-driven fusion metabolic network model and fermentation process mechanism analysis according to claim 1, wherein the genome-scale metabolic model is iFX1172 model.
9. A method of optimizing a fermentation process using the system of any one of claims 1-8, comprising executing the strategy generated by the optimization strategy output module: sodium acetate is exogenously supplemented at the rate of 0.015-0.2 g/L/h, and is added with glucose in a co-feeding mode in the middle and later stages of fermentation.
10. A method of metabolic engineering using the system of any one of claims 1-8, comprising executing a strategy generated by an optimization strategy output module: Overexpressing one or more genes of acetyl-CoA synthetase, adenosine kinase, adenosine homocysteine enzyme, glucosamine-1-phosphate N-acetyltransferase, UTP-glucose-1-phosphate uridyltransferase, glutamine synthetase, phenylalanine aminotransferase; and/or knocking out or down one or more genes of glucose phosphomutase, pentose phosphate mutase, pyruvate phosphodikinase, glutamine phosphoryl diphosphate aminotransferase, glucose-6-phosphate dehydrogenase, citrase synthase, pyridoxal 5-phosphate synthase, malonyl-coA-ACP transacylase.

Description

Collaborative optimization system based on data-driven fusion metabolic network model and fermentation process mechanism analysis Technical Field The invention relates to the technical field of fermentation engineering, in particular to a collaborative optimization system based on data-driven fusion metabolic network model and mechanism analysis in a fermentation process. Background Conventional fermentation optimization strategies rely mainly on empirical trial and error methods or the adjustment of a single physiological parameter, which often lack systemicity, and it is difficult to accurately resolve complex metabolic networks involved in gentamicin biosynthesis. Since secondary metabolism is regulated at multiple levels of metabolism, transcription, and translation, the resolution of the regulatory mechanism is often limited by the complexity of experimental conditions and the lack of data acquisition. At present, a genome scale metabolic model (GEM) becomes an important tool for analyzing a microbial metabolic network, and the GEM can integrate multiple groups of chemical data, reconstruct a global metabolic network of the microorganism and provide theoretical guidance for metabolic engineering and fermentation process optimization. Meanwhile, the ML technology has natural advantages in the aspect of processing complex, high-dimensional and nonlinear biological system data, and can identify key factors influencing product synthesis from large-scale experimental data and optimize process parameters. However, the existing researches focus on single application of GEM or ML methods, and complementary advantages between the two are not fully exerted, so that a collaborative optimization system based on a data-driven fusion metabolic network model and mechanism analysis of a fermentation process is provided. Disclosure of Invention The invention aims to provide a collaborative optimization system based on a data-driven fusion metabolic network model and mechanism analysis in a fermentation process, so as to solve the problem of constructing a GEM-ML collaborative optimization framework in the background technology. In order to achieve the purpose, the invention provides the following technical scheme that the collaborative optimization system based on the data-driven fusion metabolic network model and the mechanism analysis of the fermentation process comprises the following components: the data acquisition and processing module is used for acquiring online process parameters and multi-source spectrum prediction data in the microbial fermentation process and constructing a fermentation process data set; the metabolism flux simulation module is used for carrying out flux balance analysis on the microbial metabolism network based on the genome scale metabolism model to obtain a metabolism flux data set; the data fusion module is connected with the data acquisition and processing module and the metabolic flux simulation module and is used for fusing the fermentation process data set and the metabolic flux data set to construct a mixed data set; The machine learning modeling and optimizing module is connected with the data fusion module and is used for constructing a target product yield prediction model by adopting a machine learning algorithm based on the mixed data set; the key node identification module is connected with the machine learning modeling and optimizing module and is used for identifying key metabolic reactions and metabolic pathways affecting target product synthesis from the prediction model by utilizing a feature importance analysis method; And the optimizing strategy output module is connected with the key metabolism node identification module and is used for generating and outputting a fermentation process optimizing strategy based on the key metabolism reaction and the metabolism path. Preferably, the data fusion module adopts a time alignment and interpolation method to match the metabolic flux data with the process parameter data on a time scale. Preferably, in the machine learning modeling and optimization module, the machine learning algorithm is an Artificial Neural Network (ANN), the input is the hybrid data set, and the output is the specific synthesis rate of the target product. Preferably, the key node identification module outputs the key metabolic reaction and its SHAP value or weight score by adopting a SHAP analysis method. Preferably, the optimization strategy output module supports a bayesian optimization or response surface method for determining an optimal feed proportioning. Preferably, the system is configured for optimizing the fermentation process of Micromonospora echinocpora. Preferably, the target product is gentamicin C1a. Preferably, the genome-scale metabolic model is the iFX1172 model. A method for optimizing fermentation process by using the system comprises the steps of executing a strategy generated by an optimization strategy output module: sodium acetate i