Scientists have successfully applied reinforcement learning to molecular biology challenges.
A team of researchers has developed a powerful new protein design software that employs strategies proven proficient in board games such as chess and Go. In one experiment, proteins made with the new approach were found to be more effective at generating useful antibodies in mice.
The findings were reported on April 21st. chemistrysuggesting that this breakthrough could soon lead to more potent vaccines. More broadly, this approach could lead to a new era in protein design.
“Our results show that reinforcement learning can do more than master board games. When trained to solve long-standing puzzles in protein science, software excels at creating useful molecules.” is a Seattle M.D. and recipient of the 2021 Life Science Breakthrough Award.
“If this method is applied to the right research questions, it could accelerate progress in various scientific fields,” he said.
This work is a milestone for conducting protein science research using artificial intelligence. The potential uses are vast, from developing more effective cancer treatments to creating new biodegradable fibers.
Reinforcement learning is a type of machine learning in which a computer program learns to make decisions by trying different actions and receiving feedback. Such algorithms can learn how to play chess. For example, by testing millions of different moves that lead to victory or defeat on the board. This program is designed to help you learn from these experiences and make better decisions over time.
To create a reinforcement learning program for protein design, scientists gave a computer millions of simple starting molecules. The software then made 10,000 attempts to randomly improve each one towards a predefined goal. Computers have learned how to lengthen or bend proteins in a certain way to contort them into desired shapes.
Baker Lab members Isaac D. Lutz, Shunzhi Wang, and Christoffer Norn led the study. Their team’s scientific manuscript is entitled “Top-Down Design of Protein Architectures with Reinforcement Learning.”
“Our approach is unique because it uses reinforcement learning to solve the problem of creating protein shapes that fit together like puzzle pieces,” explains co-author Lutz. . “This is simply not possible with previous approaches and could change the types of molecules that can be built.”
As part of this research, scientists manufactured hundreds of AI-designed proteins in the lab. Using electron microscopy and other instruments, they confirmed that many of the computer-generated protein shapes were actually realized in the laboratory.
“This approach proved to be not only accurate, but also highly customizable. For example, we asked the software to create spherical structures without holes, small holes, or large holes. The possibility of creating all kinds of architectures has not yet been fully explored,” said co-author Shunzhi Wang, a postdoctoral fellow at the UW Institute of Medicine.
The team focused on designing new nanoscale structures composed of many protein molecules. This required designing both the protein components themselves and the chemical interfaces that allow self-assembly of the nanostructures.
Electron microscopy confirmed that numerous AI-designed nanostructures can be formed in the laboratory. As a measure of how accurate the design software has become, scientists have observed many unique nanostructures in which all atoms were found to be in their intended locations. In other words, the deviations between the intended and realized nanostructures were on average less than a single atom wide. This is called an atomically accurate design.
The authors foresee a future where this approach will enable the creation of therapeutic proteins, vaccines, and other molecules that could not be created by previous methods.
Researchers at the UW Medicine Institute for Stem Cell and Regenerative Medicine used primary cell models of vascular cells to show that engineered protein scaffolds outperform previous versions of the technology. For example, the receptors that help cells receive and interpret signals were more densely packed in more compact scaffolds, and thus were more effective at promoting vascular stability.
Hannele Ruohola-Baker, professor of biochemistry at the University of Washington School of Medicine and one of the study’s authors, said of the implications of research on regenerative medicine: Diabetes, brain injury, stroke, and other situations where blood vessels are at risk. provide a new way to modulate the processes of aging and aging. ”
This work was funded by the National Institutes of Health (P30 GM124169, S10OD018483, 1U19AG065156-01, T90 DE021984, 1P01AI167966). Open Philanthropy Project and Protein Design Institute’s Audacious Project. Novo Nordisk Foundation (NNF170C0030446); Microsoft; and Amgen. Part of the research was conducted at the Advanced Light Source, a national user facility operated by Lawrence Berkeley National Laboratory on behalf of the Department of Energy.
The news release was written by Ian Haydon of the UW Medicine Institute for Protein Design.
