Dresden 2020 – wissenschaftliches Programm

Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...

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

BP: Fachverband Biologische Physik

BP 33: Protein Structure and Dynamics

BP 33.8: Vortrag

Donnerstag, 19. März 2020, 12:00–12:15, ZEU 250

Comparison of continuous and discrete Markov models of biomolecular dynamics — •Benjamin Lickert and Gerhard Stock — Universität Freiburg

Motions of biomolecular systems, recorded by molecular dynamics simulations, are often modeled as Markov processes. A very popular approach is given by Markov state models where the conformational space is divided into different states [1]. To be Markovian, the intrastate dynamics need to be significantly faster than the interstate dynamics. On the other hand, the observed dynamics can be modeled as a continuous diffusive process, called Langevin dynamics, on some low-dimensional free energy landscapes F(x^→). In this case, Markovianity is given if the system, i.e., x^→(t), evolves substantially slower than the neglected degrees of freedom, i.e., the bath surrounding the system. Recently, a data-driven approach was formulated to estimate such a Langevin model from a given trajectory x^→(t) [2]. Here, we compare the features of both modeling frameworks. While Markov state models are very appealing due to their clearly structured generation and interpretation, Langevin dynamics have the advantage that they allow for the estimation of continuously defined observables, like free energy and autocorrelations. Using molecular dynamics simulations of systems with varying complexity we have a look at these points in practice.
[1]: J.H.Prinz et al., J.Chem.Phys. 134, 174105 (2011)
[2]: N.Schaudinnus et al., J.Chem.Phys. 145, 184114 (2016)