Symmetry-Driven Discovery of Dynamical Variables in Molecular Simulations

Jeet Mohapatra, Nima Dehmamy, Csaba Both, Subhro Das, Tommi Jaakkola

Keywords: AI4Science, Molecular Dynamics, Second Order Methods, Symmetry Discovery

TL;DR: we relate effective degrees of freedom to symmetries and use symmetry discovery find important DoF such as dihedral angles in alanine dipeptide

Abstract: Molecular dynamics simulations are crucial for understanding complex biomolecular systems, but they are often hindered by the high dimensionality of the configurational space. This paper introduces two novel approaches for discovering effective degrees of freedom (DoF) in molecular dynamics simulations by leveraging approximate symmetries of the energy landscape. We present a scalable symmetry loss function compatible with existing force-field frameworks and a Hessian-based method efficient for smaller systems. Both approaches enable systematic exploration of conformational space by connecting structural dynamics to energy landscape symmetries. Applied to alanine dipeptide, our methods comprehensively sample the Ramachandran plot, including shallow minima. Simulations initiated from our DoF-sampled points converge to all important conformations, demonstrating the methods’ effectiveness in navigating complex energy landscapes. These approaches offer powerful tools for efficient exploration in molecular simulations, with potential applications in protein folding and drug discovery.

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 3296