Protein shells designed using AI were free of defects
Shutterstock/Sergey Drozd
AI can design extremely dense protein shells that could eventually lead to more potent vaccines.
The viral genetic material is housed in a protein shell. Similar lab-made shells are used in vaccines, encapsulating molecules that provoke an immune response in the body. depends on. Any defect, no matter how small, can make it less effective, destabilize it, and cause unpredictable responses in cells.Isaac Lutz and his colleagues at the University of Washington in Seattle say that using artificial intelligence and wanted to see if the design and creation of these shells could be more accurate.
They first gave the AI the properties they wanted the shell to have, such as size and porosity. Then the AI used reinforcement learning. It’s the same iterative process AI systems use to learn to play games like chess: try different moves, get feedback, and try again. Here the AI move was to bind, fold and entangle small protein structures called alpha helices into a 20-sided shell and then check if the design had the desired properties.
After AI designed hundreds of thousands of shells, researchers created about 350 shells in the lab. Examining them under an electron microscope shows that AI has created denser shells than ever created. Lutz says this is because it started with very small building blocks that could be put together more cleanly than the larger protein structures the researchers had previously used.
“Before, it seemed like you had to buy something like protein Lego first, which limited the ways you could build it. We design and connect the exact LEGOs needed to achieve the,” he says.
To test how high density affects shell usage in living cells, the team sprinkled shells with different molecules and inserted them into mice. In particular, in one experiment with a molecule that causes the production of influenza antibodies, the AI-designed shell produced a small but statistically significant immune response compared to several conventional vaccine candidates currently in clinical trials. resulted in an increase in Lutz attributed this to the accuracy of the AI methods. All molecules are placed exactly where they are needed on the shell, and the shell has enough structure to support many of them.
“It’s amazing that the team was able to achieve this. It takes billions of years of evolution to design a single protein that folds properly, but this is another level of complexity, It folds and binds proteins very well to create a closed structure.
Yang Zhang of the University of Michigan says that in addition to vaccines, AI-engineered protein shells could be useful in gene therapy, where genetic material can be placed inside a tailored shell that prevents patient cells from reacting adversely. said.
topic:
