However, the antibodies produced by RFdiffusion are far from clinical use. The designer antibodies that worked did not bind strongly to their targets and would need their sequences modified to resemble natural human antibodies to avoid provoking an immune reaction. The designs are also single-domain antibodies, which are simpler than the more complex proteins that most approved antibody drugs are based on. Despite these challenges, the researchers believe this is a significant step towards designing antibody drugs at the touch of a button.
Key takeaways:
- Researchers have used generative artificial intelligence (AI) to create new antibodies for the first time, potentially bringing AI-guided protein design to the therapeutic antibody market.
- The AI tool used, RFdiffusion, was modified to design antibodies that recognize specific regions of several bacterial and viral proteins, including those that the SARS-CoV-2 and influenza viruses use to invade cells.
- The success rate of the AI-designed antibodies was about one in 100, which is lower than the success rate achieved with other types of AI-designed protein.
- Despite the initial success, the antibodies produced by RFdiffusion are a long way from clinical use. They need to be modified to resemble natural human antibodies and are currently based on simpler single-domain antibodies found in camels and sharks.