(604d) Probing the Kinetics and Mechanism of Protein Folding from Path Sampling Simulations

Protein folding has long captivated the attention of biophysicists, with early efforts focused on predicting a protein’s three-dimensional structure from its amino acid sequence. While recent breakthroughs in machine learning—culminating in AlphaFold’s 2024 Nobel Prize-winning achievement—have largely resolved this structural prediction challenge, the problem of understanding protein folding kinetics and mechanisms remains an open challenge. Folding is a rare, activated process characterized by large free energy barriers and long latency times that are often inaccessible to conventional molecular dynamics (MD) simulations. Experimental approaches also struggle to provide detailed, atomistic insight into folding pathways due to their insufficient spatiotemporal resolution.

In this work, we address these challenges using jumpy forward flux sampling (jFFS), an advanced path sampling algorithm developed in our group, in conjunction with a scalable, GROMACS-compatible software platform. We apply this framework to investigate the folding kinetics of a fast-folding miniprotein. The computed folding times are in excellent agreement with experimental measurements, validating the accuracy of our approach. Beyond kinetics, we leverage machine learning techniques to analyze thousands of folding trajectories generated by jFFS. This analysis reveals two dominant folding pathways, distinguished by widely diffferent sequences of secondary structure formation. By quantifying the probability of each pathway, we also estimate pathway-specific folding times.

These results highlight the power of path sampling techniques—particularly jFFS—for uncovering the mechanisms and kinetics of biomolecular rare events, offering a generalizable approach for studying complex folding landscapes that are otherwise difficult to access.