Regensburg 2025 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 34: Statistical Physics in Biological Systems II (joint session DY/BP)

BP 34.4: Vortrag

Freitag, 21. März 2025, 12:30–12:45, H43

Position-Dependent Non-Markovian Effects Improve Protein Folding Simulations — •Lucas Tepper, Cihan Ayaz, Benjamin Dalton, and Roland Nezt — Freie Universität Berlin

It's common to project a protein's full atomic resolution onto a one-dimensional reaction coordinate to capture key aspects of its folding process. As a direct consequence of this dimensionality reduction, non-Markovian memory effects emerge. Accounting for memory effects in the framework of the generalized Langevin equation (GLE) with linear friction has proven efficient, accurate and insightful. However, recent advances in deriving GLEs with non-linear, position-dependent friction kernels raise questions about their applicability to protein folding simulations. We derive a novel method to extract position-dependent friction kernels from time series data via conditional Volterra equations. When applied to two protein test systems, the position- and time-dependent friction is strongest for long memory times in the folded states, where atoms are tightly packed. Additionally, we propose a novel and numerically efficient GLE simulation setup, confirming the accuracy of the extracted kernels. Compared to linear friction GLE simulations, our results show that position-dependent non-Markovian effects are critical for accurately reproducing protein folding kinetics when using low-dimensional reaction coordinates.

Keywords: Generalized Langevin Equation; Protein Folding; Position-Dependent Memory Effects