A research team, from the Washington University School of Medicine, in the United States, developed a powerful protein design software, based on a proven strategy in board games such as chess, which served to manufacture hundreds of these substances in an experiment later.
“The proteins created with the new approach were more effective in generating useful antibodies in rodents”, according to the authors of the research published in the journal Science, who suggest that “this advance may soon lead to more powerful vaccines”.
Overall, the method could usher in a new era in protein design, the university said in a statement, quoted this Thursday by the Efe agency.
“Our results demonstrate that reinforcement learning can do more than just master board games,” said study lead author David Baker.
“If this method is applied to the right research problems, it could accelerate progress in many fields,” he added.
The potential applications are vast, from developing more effective cancer treatments to creating new biodegradable tissues.
In this reinforcement learning, a program learns to make decisions by trying different actions and receiving ‘feedback’.
This algorithm can, for example, learn to play chess, trying millions of different moves that lead to victory or defeat on the board. The program is designed to learn from these experiences and improve, over time, in decision making.
In this case, the scientists gave the computer millions of starting simple molecules, and then the program made 10,000 attempts to randomly improve each one toward a predefined goal: the computer stretched the proteins or folded them in specific ways until it learned to give them the desired shape.
“Our approach is unique because we use reinforcement learning to solve the problem of creating protein shapes that fit together like pieces of a jigsaw puzzle”, stressed Isaac D. Lutz, another of the researchers.
“This was simply not possible with previous approaches and has the potential to transform the types of molecules we can build,” he added.
As part of this study, scientists produced hundreds of Artificial Intelligence (AI) designed proteins in the lab. Using electron microscopes and other instruments, they confirmed that many of the forms of proteins created by the computer became reality in the laboratory.
“This method proved to be not only accurate, but also highly customizable. For example, we asked the program to create spherical structures with no holes, with small holes or with large holes; its potential for generating all kinds of architectures has yet to be explored. “, detailed Shunzhi Wang.
The team focused on designing new nanoscale structures composed of many protein molecules. For this, it was necessary to design both the protein components themselves and the chemical interfaces that allow the self-assembly of nanostructures.
Electron microscopy confirmed that numerous AI-engineered nanostructures were capable of being formed in the laboratory.
To measure the accuracy of the design software, the scientists analyzed many unique nanostructures in which every atom was in its intended place.
The deviation between the predicted and realized nanostructure was, on average, less than the width of a single atom.
This is what is called, the authors point out, “atomically precise design”.
Read Also: Artificial Intelligence recreated how a new album by the band Oasis would sound
Always be the first to know.
Seventh consecutive year Consumer Choice for Online Press.
Download our free App.
