The new tool, ProteinMPNN, described by a group of researchers from the University of Washington in two articles published in Science today (available here other here), offers a powerful complement to that technology.
The papers are the latest example of how deep learning is revolutionizing protein design by giving scientists new research tools. Traditionally, researchers design proteins by modifying those that occur in nature, but ProteinMPNN will open up a whole new universe of possible proteins for researchers to design from scratch.
“In nature, proteins basically solve all of life’s problems, from harvesting energy from sunlight to making molecules. Everything in biology happens from proteins,” says David Baker, one of the scientists behind the article and director of the Institute for Protein Design at the University of Washington.
“They evolved throughout evolution to solve problems that organisms faced during evolution. But today we are facing new problems, 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, that would be very, very powerful.”
Proteins consist of hundreds of thousands of amino acids that are linked together in long chains, which are then folded into three-dimensional shapes. AlphaFold helps researchers predict the resulting structure, providing insights into how they will behave.
ProteinMPNN will help researchers with the reverse problem. If you already have an exact protein structure in mind, it will help you find the sequence of amino acids that folds into that shape. The system uses a neural network trained on a large number of exemplary amino acid sequences, which fold into three-dimensional structures.
But the researchers also need to solve another problem. To design proteins that tackle real-world problems, such as a new plastic-digesting enzyme, they first have to figure out what protein structure would do the job.
To do this, researchers in the Baker lab use two machine learning methods, detailed in a Article in Science last July, which the team calls “restricted hallucination” and “in painting.”