AI-supported metabolite analysis for the control of biotechnological fermentation processes

The global food situation is facing profound changes: climate change, pollution, limited fertilizer resources, geopolitical developments, and increasing demands for environmental and sustainability standards are putting conventional agriculture under growing pressure. At the same time, the demand for climate-friendly, locally produced, and sometimes vegan foods is increasing. To ensure long-term food security, new food sources (novel food) are therefore coming into focus. These include previously unused plant and animal species, innovative processing technologies, and products manufactured using specialized production organisms—such as microorganisms, filamentous fungi, yeasts, or plant cell cultures—in biotechnological processes. These organisms enable resource-efficient, circular production methods and open up new opportunities for high-quality foods and functional ingredients.

However, bioreactor-based production processes are complex and costly. Besides established, classic substrates like glucose or glycerol, there is great potential in utilizing more heterogeneous substrates from residual material streams in the future – that is, raw material mixtures whose composition (e.g., sugar, protein, fat, or fiber content) is subject to natural fluctuations. Currently, however, such by-products are only used to a limited extent due to their variable properties and regulatory frameworks. In the long term, however, they could contribute significantly to resource conservation – provided that stable and efficient processes can be implemented despite this variability.

The control technology developed for classic fermentation processes quickly reaches its limits with this combination of variable feedstocks, numerous interacting process parameters, and dynamic biological systems. As a result, the potential of modern biotechnological processes for the production of new foods remains largely untapped. MiKI addresses precisely this issue and develops data-driven, more intelligent approaches to both improve established biotechnological production processes and facilitate the future use of more sustainable substrate sources.

The goal of MiKI is to develop a novel, data-driven, and AI-supported control technology that makes complex bioreactor-based production processes for novel foods significantly more robust, efficient, and economical. The focus is on the ability to precisely control biotechnological processes even when classical models fail and process parameters are difficult to predict. A key element of the approach is continuous, real-time metabolic monitoring using high-resolution GC-IMS analysis (ion mobility spectrometry with gas chromatographic pre-separation). Metabolite profiles obtained from the process exhaust air provide a dynamic and highly detailed picture of the biological process state. This data forms the basis for an AI-supported correlation analysis that identifies relationships between process parameters, standard sensors, and metabolite patterns.

Through statistically planned experiments (DoE), predictive, model-independent ("black-box") AI models are derived from this data, enabling reliable predictions of the process progression. Building on this foundation, an intelligent, adaptive process control system is being developed that responds specifically to changes in culture behavior and optimizes the process with regard to yield, stability, and resource efficiency. To demonstrate the transferability of the approach, the technology is being developed and evaluated in three exemplary production systems:

• Yeast (Komagataella phaffii) for the production of a protein-based sugar substitute from the "NovelSweets" and "BeyondSugar" projects;
• filamentous fungus for the production of biomass-based novel food components
• plant cell culture (hop cell lines) for the extraction of plant secondary metabolites for food applications.
  

Fraunhofer IME is responsible for the overall coordination of the project and for key tasks related to the development and implementation of the intelligent process control system, as well as the transfer of the technology to the aforementioned example processes.