SFB 1551 Seminarreihe Veranstaltung - Dr. Oleksandra Kukharenko

Methodenvortrag von Dr.Olksandra Kukharenko, Max-Planck-Institut für Polymerforschung

Title: Bridging Scales in Molecular Dynamics Simulations with Machine Learning

Zusammenfassung:
Grasping the behavior of biomolecules across diverse length and time scales is a persistent challenge in molecular biophysics. While atomistic simulations provide detailed molecular insights, their high computational cost limits their applicability. Coarse-grained models, on the other hand, allow access to longer timescales and larger systems but often sacrifice resolution and accuracy. To connect these approaches meaningfully, it is essential to develop strategies that can bridge the scales and preserve key thermodynamic properties.

In this talk, I will showcase recent advances in leveraging machine learning for the analysis of free energy landscapes. I will demonstrate how tools such as dimensionality reduction and clustering can uncover key collective variables and metastable states from high-dimensional simulation data. Moreover, I will explore how these learned representations can facilitate the transfer of insights between different modeling resolutions, enabling seamless integration of coarse-grained and atomistic simulations. Applications to systems with complex multi-body interactions will illustrate the power of this multiscale, data-driven perspective.