Öffnet in neuem Fenster Opens in a new window Öffnet externe Seite Opens an external site Öffnet externe Seite in neuem Fenster Opens an external site in a new window

Department Optics and Beamlines

AI-based enzyme design

In this project we use literature data mining to link enzyme sequence and activity data to train novel diffusion models for enzyme design. The AI-generated enzyme sequences will be tested for activity in vitro to validate the prediction accuracy of the trained models. A first proof of concept is planned with the ß-lactamase enzyme. The framework will later be applied to PETase to allow for optimization of plastic degrading enzymes.

260120 enzyme design project

Data mining for enzyme sequences and activity are the basis for the training of novel diffusion models to predict new enzyme structures.

This project started in 01/2026 together with Dr. Gert Weber of the Research Group Macromolecular Crystallography. 

Contribution from ZIB: 

  • Chatbot hosted at ZIB generated code and supported literature research
  • HPC ressources are planned to be used in model training and application

© 01/26 Yannic Kerkhoff