Protein-designing AI could discover new cures and materials unknown to science

A group of researchers from the University of Washington has engineered a new AI tool that can identify and design new proteins.

This could lead to more efficient vaccines, better cures for cancer, or new materials, according to a report published by MIT Technology Review on Thursday.

Using proteins to solve nature's problems

The new tool is called ProteinMPNN and is described in two papers in Science journal, available here and here.

“In nature, proteins solve basically all the problems of life, ranging from harvesting energy from sunlight to making molecules. Everything in biology happens from proteins,” said David Baker, director of the Institute for Protein Design at the University of Washington.

“They evolved over the course of evolution to solve the problems that organisms faced during evolution. But we face new problems today, like covid. If we could design proteins that were as good at solving new problems as the ones that evolved during evolution are at solving old problems, it would be really, really powerful,”added Baker, who is also a co-author of the paper.

ProteinMPNN will help researchers that have a protein structure in mind to find the amino acid sequence that makes up its exact shape. It does this by employing a neural network trained on a very large number of examples of amino acid sequences.

However, in order to ensure the researchers design proteins that are useful for real-world applications, they first have to establish what protein backbone would have the function they are seeking to activate.

To achieve that, they use two methods called “constrained hallucination,” which lets users do a random search among all possible protein sequences, and “hallucination,” allowing them to explore the space of all possible protein structures.

Going beyond the proteins found in nature

These two processes allow researchers to go beyond the proteins merely found in nature and create entirely new ones. “Nature has only sampled … a tiny fraction. So if you limited the search to those sequences that exist in nature, you wouldn’t get anywhere,” Baker said.

The new software is more than 200 times faster compared to similar previous best tools. "These contributions and others recently are transforming the field of biomolecular structure prediction and design,” said Jeffrey Gray, a professor of chemical and biomolecular engineering at Johns Hopkins University. 

“The implications are dramatic in terms of understanding biology, health, and disease and in designing new molecules to reduce human suffering,” Gray added.

Best of all, ProteinMPNN is now available for free on the open-source software repository GitHub. The tool will now allow scientists to make unlimited new protein designs for many useful current applications.

“The challenge, of course … is what are you going to design?” Baker said. Indeed, it seems that technology has finally caught up to the possibilities of the imagination of scientists. With this new tool, researchers can construct whatever protein they can conceive of, and that is something to get excited about.