Arizona State University Researchers Unlock Hidden Protein Rhythms to Accelerate Drug Discovery and Machine Learning Models

ASU researchers use supercomputing to map protein rhythms in under 24 hours. Learn how "sequence-to-dynamics" is transforming drug design and AI in 2026.

By: AXL Media

Published: Mar 28, 2026, 4:21 AM EDT

Source: Information for this report was sourced from Arizona State University

Decoding the Rhythms of Molecular Motion

Proteins are far more than static structural components, they are dynamic biomolecules that must shift between multiple shapes to drive metabolic reactions and immune responses. For years, the scientific community struggled to predict these transitions because traditional tools were designed for rapid, minute vibrations rather than the slow, sweeping shifts that define protein function. According to Associate Professor Matthias Heyden of ASU, proteins move according to deep rhythms, similar to a skyscraper swaying in the wind. By identifying these low-frequency patterns, researchers can now predict how a protein will bend and twist without needing to observe the full motion in real time.

Accelerating Simulations Through Molecular Collisions

The innovation from the Heyden research group involves a method that extracts meaningful data from computer snapshots lasting only billionths of a second. The team resurrected the theory that major shape changes are fundamentally tied to natural fluctuations caused by molecular collisions at room temperature. According to the study published in Science Advances, analyzing these tiny, ever-present tremors allows scientists to identify the "path of least resistance" for a protein. This is compared to testing a door to see if it pushes or pulls, a small nudge reveals the intended direction of travel without requiring the door to be fully opened.

From Static Structures to Dynamic Ensembles

While recent advancements like AlphaFold have revolutionized the ability to predict a protein's static shape from its genetic sequence, the ASU team is expanding this into the realm of dynamics. By using the identified vibrations as guide rails, the researchers nudged five distinct proteins along their natural pathways to map their energetic landscapes. This approach allows for the high-throughput generation of "conformational ensembles," which are essentially catalogs of every shape a protein prefers to adopt. This shift from "sequence-to-structure" to "sequence-to-structure-to-dynamics" is expected to provide the rich datasets needed for the next generation of machine learning in biology.