University of Washington graduates use artificial intelligence to create new proteins

SEATTLE, Wash.-

For over two years, protein structure prediction has been changed by machine learning. On Sept. 15, two science related research talk about a similar idea in the revolution of protein design. 

The findings show how machine learning can create protein molecules that are more accurate and made quicker than before. 

“With these new software tools, we should be able to find solutions to long-standing challenges in medicine, energy, and technology,” said senior author David Baker, professor of biochemistry at the University of Washington School of Medicine.  

The algorithm used in machine learning which includes RoseTTAFold have been trained to predict the smaller detailed shapes if natural proteins based on their amino acid sequences. 

Machine learning is a type of artificial intelligence that allows computers to learn from data without having to be programmed. 

A.I. has the ability to generate protein in two ways. One being akin to DALL-E or other A.I. tools that produce an output from simple prompts. The second is the autocomplete feature we can find in a search bar. 

As a way of making things go by faster the A.I. team created a new algorithm that creates amino acid sequences. This tool, called ProteinMPNN, creates the sequence in one second. That's over 200 minutes faster than previous best software. 

The Baker Lab also says combining new machine learning tools could reliably generate new proteins that functioned in the laboratory. Among those were the nanoscale ring that could make up part of a custom nanomachines.