Research Highlight: Atomically accurate de novo design of antibodies with RFdiffusion
Antibodies are among the most important classes of therapeutic molecules, but designing antibodies with precise structures and binding properties remains a major challenge. Traditional antibody discovery relies heavily on experimental screening or modifying existing antibody scaffolds, which can limit the exploration of new structures and functions.
In this talk, Nate Bennett will present recent advances in AI-driven protein design that enable the de novo generation of antibodies with atomic-level accuracy. Using RFdiffusion, a deep-learning framework for generative protein design, researchers can computationally design entirely new antibody structures that match target geometries and binding constraints.
The approach integrates structure-based generative modeling with protein design principles, allowing the creation of antibodies that precisely target defined epitopes. Experimental validation demonstrates that these computationally designed antibodies can fold correctly and bind their intended targets.
This work highlights how generative AI methods are transforming antibody discovery, enabling the rapid design of new therapeutic molecules and expanding the possibilities for programmable protein engineering.
About the speaker
Nate Bennett
Nate Bennett is a co-founder of Xaira Therapeutics, a biotechnology company developing AI-driven approaches for therapeutic discovery.
He completed his PhD at University of Washington, where he worked on computational protein design and generative modeling for biomolecular engineering. Following his doctorate, he continued at the university as a postdoctoral scholar before co-founding Xaira Therapeutics.
Nate previously earned a BSc in Chemical Engineering from University of Wisconsin–Madison, where he worked on metabolic and microbial engineering projects. His research focuses on developing machine learning methods for protein design, with applications in antibody engineering and therapeutic discovery.