Abstract:

Computational protein design is emerging as a powerful tool to create proteins with novel or enhanced functionalities. However, most designed proteins to date are composed of structurally simple topologies, far from the complexity sampled in nature. To overcome this limitation, we developed a deep learning-based pipeline leveraging the incredible accuracy of AlphaFold2 for the design of proteins with complex natural protein topologies and high experimental success rates. We applied our approach to the design of soluble analogues of membrane proteins, such as GPCRs and claudins. We demonstrate that our soluble analogues are highly stable, structurally accurate, and are able to support native epitopes for antibody or G-protein binding in solution. We then extended the capabilities of our pipeline to the design of highly specific protein binders. We are now able to design binders with unprecedented experimental success rates against therapeutically relevant targets, such as PD-L1 or CD45, as well as much more challenging targets, such as CRISPR-Cas nucleases, Argonautes, and common allergens. These advancements pave the way for the accurate design of proteins with complex functions and potential applications in research, biotechnology, and therapies.

PhD students and postdocs are invited to join the lunch with Dr. Martin Pacesa at 12:00-13:30. Please inform Patrick Pausch if you’ll be joining for lunch 

From 14:00 to 17:00, there will be available slots for individual meetings with Dr. Martin Pacesa. Please reach out to Patrick to reserve a time: 

We are looking forward to welcome you at the seminar on April 1!