Lecture
Digital tools to discover and characterize carbohydrate-active enzymes for producing functional food ingredients
- at -
- ICM Saal 4b
- Type: Lecture
Lecture description
D. Wefers, Halle/DE1, L. Ernst, Halle/DE1, F. Stadler, Halle/DE1, F.E. Herrera Rocha, Halle/DE2, L. Pannier, Halle/DE1, M. D. Davari, Halle/DE2,
1 Martin Luther University Halle-Wittenberg, Institute of Chemistry, Food Chemistry
2 Leibniz Institute of Plant Biochemistry, Computational Chemistry
The enzymatic conversion of carbohydrates has great potential for producing functional food ingredients. For example, different sucrase enzymes enable the conversion of sucrose into non-digestible carbohydrates, including dextran and other complex α-glucans, as well as the fructans levan and inulin. As the underlying reactions use sucrose as the only substrate, they can be carried out in simple sucrose solutions or even directly in sucrose-containing food products. In addition, hydrolytic carbohydrateactive enzymes can be used to cleave the polymers into oligosaccharides which exhibit different functionalities. However, in many cases, only a few enzymes are known, and these often exhibit non-ideal properties. Novel enzymes can be identified by using the
sequences of known enzymes as queries in global sequence alignment analyses. However, such homology-based approaches are inherently biased toward the identification of enzymes closely related to already known sequences, thereby limiting the discovery of more divergent or novel functionalities. The limited diversity of the resulting sequences can further impede rational enzyme engineering aimed at improving their properties. The machine-learning assisted mining of enzyme sequences supported by homology modelling can overcome these limitations [1]. An integrative approach based on the SelectZyme enzyme mining tool, homology modeling, structural alignment with the available crystal structure of the enzyme family, and substrate docking was applied to identify novel endo-levanase candidates for the hydrolysis of levan into oligosaccharides with potential application as sugar substitutes. The optimization of carbohydrate conversion by new and previously described enzymes also remains a complex task, as multiple parameters such as temperature, pH, enzyme activity, substrate concentration, and incubation time have to be considered. The large number of possible parameter combinations makes exhaustive experimental testing impractical and necessitates structured experimental designs. While statistical design of experiments can address this challenge, predictive algorithms offer increased efficiency. Although open-source algorithms such as Posthog or Unleash are available, they are not optimized or directly applicable for the analysis of enzymatic reactions. Therefore, the MixMOBO algorithm [2] was adapted and implemented in MATLAB with a graphical user interface. Overall, the described digital tools show strong potential to support the development of new food ingredients based on enzymatic reactions.
Literature:
[1] F. Moorhoff, Y. Zhang, S. Qiu, W. Dong, D. Medina-Ortiz, J. Zhao, M. D. Davari, ACS Catal., 2026, 16, 12–30.
[2] H. M. Sheikh, P. S. Marcus, Struct. Multidiscip. Optim., 2022, 65, 331.
1 Martin Luther University Halle-Wittenberg, Institute of Chemistry, Food Chemistry
2 Leibniz Institute of Plant Biochemistry, Computational Chemistry
The enzymatic conversion of carbohydrates has great potential for producing functional food ingredients. For example, different sucrase enzymes enable the conversion of sucrose into non-digestible carbohydrates, including dextran and other complex α-glucans, as well as the fructans levan and inulin. As the underlying reactions use sucrose as the only substrate, they can be carried out in simple sucrose solutions or even directly in sucrose-containing food products. In addition, hydrolytic carbohydrateactive enzymes can be used to cleave the polymers into oligosaccharides which exhibit different functionalities. However, in many cases, only a few enzymes are known, and these often exhibit non-ideal properties. Novel enzymes can be identified by using the
sequences of known enzymes as queries in global sequence alignment analyses. However, such homology-based approaches are inherently biased toward the identification of enzymes closely related to already known sequences, thereby limiting the discovery of more divergent or novel functionalities. The limited diversity of the resulting sequences can further impede rational enzyme engineering aimed at improving their properties. The machine-learning assisted mining of enzyme sequences supported by homology modelling can overcome these limitations [1]. An integrative approach based on the SelectZyme enzyme mining tool, homology modeling, structural alignment with the available crystal structure of the enzyme family, and substrate docking was applied to identify novel endo-levanase candidates for the hydrolysis of levan into oligosaccharides with potential application as sugar substitutes. The optimization of carbohydrate conversion by new and previously described enzymes also remains a complex task, as multiple parameters such as temperature, pH, enzyme activity, substrate concentration, and incubation time have to be considered. The large number of possible parameter combinations makes exhaustive experimental testing impractical and necessitates structured experimental designs. While statistical design of experiments can address this challenge, predictive algorithms offer increased efficiency. Although open-source algorithms such as Posthog or Unleash are available, they are not optimized or directly applicable for the analysis of enzymatic reactions. Therefore, the MixMOBO algorithm [2] was adapted and implemented in MATLAB with a graphical user interface. Overall, the described digital tools show strong potential to support the development of new food ingredients based on enzymatic reactions.
Literature:
[1] F. Moorhoff, Y. Zhang, S. Qiu, W. Dong, D. Medina-Ortiz, J. Zhao, M. D. Davari, ACS Catal., 2026, 16, 12–30.
[2] H. M. Sheikh, P. S. Marcus, Struct. Multidiscip. Optim., 2022, 65, 331.
