When AI Discovers the Physics of Proteins: New Publication in Physical Review Letters

When AI Discovers the Physics of Proteins: New Publication in Physical Review Letters
Professor Qianyuan Tang and his team reveal universal physical constraints hidden in AI-predicted protein structures

Artificial intelligence (AI) can now predict protein structures with remarkable accuracy, yet what these models actually learn about the physical principles governing proteins has remained largely unknown. A new study led by researchers from the Department of Physics at HKBU shows that protein-structure AI models implicitly encode a unifying physical constraint linking protein folding topology, molecular dynamics, and evolutionary organization. The findings were recently published in Physical Review Letters.

By revealing the fundamental physical rules that govern protein behavior, the study provides insights central to therapeutic protein engineering, understanding disease-related misfolding, and designing enzymes for industrial applications. These insights could reshape how researchers approach drug design. More broadly, the work illustrates an emerging "AI for Science" paradigm in which interpretable physical laws are extracted by combining AI-generated data with statistical-physics analysis, turning prediction tools into engines of scientific discovery.

The study was led by Professor Qianyuan Tang of HKBU’s Department of Physics, with RPg student Zecheng Zhang (HKBU) as first author and RPg student Yuxiang Zheng(HKBU) contributing significantly. Collaborating researchers are from the Wenzhou Institute, University of Chinese Academy of Sciences, Jiangsu University of Technology, and Nanjing University.

We extend our warm congratulations to Professor Tang and his team on this prestigious publication, which brings fresh insights to the scientific community and highlights how physics and AI are jointly advancing our understanding of the mechanics of life.

DOI: https://doi.org/10.1103/7j2j-f8f7