Scientists are turning the process of protein folding into sound

Scientists have been studying how proteins fold for decades. This research got a boost in late 2020, when AlphaFold proved it could predict protein structures with great accuracy. But what exactly happens as proteins transition from unfolded to folded states and vice versa? This is information that scientists have struggled to visualize. So University of Illinois Urbana-Champaign chemistry researcher Martin Gruebele and composer and software developer Carla Scaletti teamed up to try sound instead.

Proteins are vital for life, involved in virtually every biochemical process that makes life possible. They comprise one or more amino acid chains (called polypeptides) that are folded into complex 3D structures. These configurations are very specific and determine a protein’s function. Misfolded proteins are responsible for conditions like Alzheimer’s disease, Parkinson’s disease, cystic fibrosis, and diabetes.

Scaletti and Gruebele mapped the hydrogen bonds—relatively weak bonds between polypeptide chains and between the chains and the surrounding watery cell environment—that formed as a protein went from an unfolded to folded state. The researchers assigned each type of bond a pitch. They also used molecular dynamics simulations and data analytics to represent the folding and unfolding events (Proc. Natl. Acad. Sci. U.S.A. 2024, DOI: 10.1073/pnas.2319094121).

When the researchers played the sounds back, they found that the proteins could cross the barrier from unfolded to folded states and vice versa in different ways. Some hydrogen bonds sped up the process, and others slowed it down. “There were not infinitely many ways you could get across, but there were certainly at least three,” Gruebele says. They classified these ways as highway, ambiguous, and meander.

In highway mode, the proteins folded really fast. In meander mode, they made initial, intermediate contacts, broke apart, and tried again till they got it right. “We’re saying faster and slower, but it’s still on the order of tens of nanoseconds,” Scaletti says.

The researchers also noticed that the amino acid residuesthat slowed down the folding process weren’t random but the same ones every time.

They classified transitions that were between highway and meander as ambiguous. This third pathway, distinct from the other two, was verified by computational studies.

“From the point of view of sonification, we’re not just interested in sound at the beginning and the end; we’re interested in how you get from the beginning to the end,” Scaletti says. Scaletti and coauthor Kurt J. Hebel developed the software the researchers used, called Kyma.

Birte Höcker, a molecular biochemist at the University of Bayreuth who wasn’t involved with the study, says that despite structure prediction’s having become much more precise in recent years, scientists still lack a detailed understanding of the steps of the process. Methods that help analyze this process are thus needed, she says. “The data analysis performed here provides very nice new aspects.”

She adds that with the sonification, “you can now hear when contacts are made, and it’s a different way of perceiving what happens during folding.”