Washington: Scientists have developed an artificial intelligence system capable of creating artificial enzymes from scratch. In laboratory experiments, some of these enzymes performed as well as those found in nature, even though their artificially created amino acid sequences differed vastly from any known natural enzymes. protein.
The experiment shows that natural language processing, although developed to read and write language text, can learn at least some of the basic principles of biology. Sales force Research developed the AI program, called progenywhich uses the prediction of the next token to assemble amino acid sequences into artificial proteins.
The scientists said the new technology could become more powerful than directed evolution, the Nobel Prize-winning protein design technology, and will energize the 50-year-old field of protein engineering by accelerating the development of new proteins that can be used to almost everything. from therapeutics to degrading plastic.
“Artificial designs work much better than designs inspired by the evolutionary process,” James said. phrasesPhD, professor of bioengineering and therapeutic sciences in the UCSF School of Pharmacy and author of the paper, which was published January 26 in Nature Biotechnology.
An earlier version of the article has been available on the preprint server BiorXiv since July 2021, where it garnered several dozen citations before being published in a peer-reviewed journal.
“The language model is learning aspects of evolution, but it’s different from the normal evolutionary process,” Fraser said. “We now have the ability to fine tune the generation of these properties for specific effects. For example, an enzyme that is incredibly heat stable or likes acidic environments or doesn’t interact with other proteins.”
To create the model, the scientists simply fed the amino acid sequences of 280 million different proteins of all kinds into the machine learning model and let it digest the information for a couple of weeks. They then fitted the model by building it with 56,000 sequences from five lysozyme families, along with contextual information about these proteins.
The model quickly generated a million sequences, and the research team selected 100 to test, based on how closely they resembled natural protein sequences, as well as how naturalistic the “grammar” and “semantics” were. “of the underlying amino acids of AI proteins.
From this first batch of 100 proteins, which Tierra Biosciences examined in vitro, the team created five artificial proteins to test in cells and compared their activity to an enzyme found in the white of chicken eggs, known as chicken egg white lysozyme. chicken egg. (TO TRY). Similar lysozymes are found in human tears, saliva, and milk, where they defend against bacteria and fungi.
Two of the artificial enzymes were able to break down the cell walls of the bacteria with activity comparable to HEWL, but their sequences were only 18 percent identical to each other. The two sequences were approximately 90 percent and 70 percent identical to any known protein.
Just one mutation in a natural protein can cause it to stop working, but in a different round of testing, the team found that the AI-generated enzymes showed activity even when as little as 31.4 percent of their sequence resembled anything. known natural protein.
The AI was even able to learn how the enzymes should be formed, simply by studying the raw sequence data. Measured with X-ray crystallography, the atomic structures of the artificial proteins looked as they should, though the sequences were unlike anything seen before.
Salesforce Research developed ProGen in 2020, based on a type of natural language programming its researchers originally developed to generate English text.
They knew from their previous work that the AI system could learn the grammar and meaning of words, along with other underlying rules that make for well-composed writing.
“When you train sequence-based models with a lot of data, they’re really powerful at learning structures and rules,” he said. Nikhil Naik, PhD, director of AI research at Salesforce Research and lead author of the paper. “They learn which words can coexist and also the composition.”
With protein, the design options were nearly limitless. Lysozymes are small like proteins, with up to about 300 amino acids. But with 20 possible amino acids, there is a huge number (20,300) of possible combinations. That’s greater than taking all the humans who have ever lived, times the number of grains of sand on Earth, times the number of atoms in the universe.
Given the limitless possibilities, it is remarkable that the model can so easily generate working enzymes.
“The ability to generate functional proteins from scratch demonstrates that we are entering a new era of protein design,” said Ali Madani, PhD, founder of Profluent Bio, former Salesforce Research Research Scientist, and the paper’s author. First author. “This is a versatile new tool available to protein engineers, and we look forward to seeing therapeutic applications.”
