Made-to-measure proteins: Nobel laureate Baker and Nvidia open AI model RFD4-Proteina
The University of Washington's Institute for Protein Design and Nvidia have unveiled an open AI model for designing new proteins. It is said to deliver five to ten times more hits.
Photo: Arthur Petron / Wikimedia Commons, CC BY-SA 4.0 (cropped and resized)
Video: YouTube · NVIDIA · youtube.com
The University of Washington's Institute for Protein Design (IPD) and chipmaker Nvidia have jointly developed the AI model RFD4-Proteina. It designs proteins with targeted functions and combines the IPD's RosettaFold Diffusion line with Nvidia's Proteina-Complexa model. What sets it apart: the model adjusts a protein's sequence and structure together with the molecules it is meant to interact with – when one part changes, the others are updated. For harder tasks it can spend more computing time.
According to the university, design success rates on computational benchmarks are five to ten times higher than with earlier models, and candidates are generated five to ten times faster. The model is freely available under the open BSD-3 licence. “RFD4-Proteina will open the door to new solutions in medicine, sustainability, materials and biotechnology,” said IPD director David Baker, who received the 2024 Nobel Prize in Chemistry.
Early in-house users report quick results: doctoral student Seth Woodbury said that “in just a few days” they designed a pesticide-degrading enzyme that performs many times better than his previous best. The project was led by doctoral students Nathaniel Corley, Maxwell Kazman and Rafi Brent.