Autodifferentiable Geometric Restraints for Enhanced Sampling Simulations with Classical and Machine Learned Force Fields
Abstract
The use of external restraints is ubiquitous in advanced molecular simulation techniques. In general, restraints serve to reduce the configurational space that is available for sampling, thereby reducing the computational demands associated with a given simulations. Examples include the use of positional restraints in docking simulations or positional restraints in studies of catalysis. Past work has sought to couple complex restraining potentials with enhanced sampling methods, including Metadynamics or Extended Adaptive Biasing Force approaches. Here, we introduce the use of more general geometric potentials coupled with enhanced sampling methods that incorporate neural networks or spectral decomposition to achieve more efficient sampling in the context of advanced materials design.
Cite
@article{arxiv.2504.13575,
title = {Autodifferentiable Geometric Restraints for Enhanced Sampling Simulations with Classical and Machine Learned Force Fields},
author = {Gustavo R. Pérez-Lemus and Cintia A. Menendez and Yinan Xu and Pablo F. Zubieta Rico and Yezhi Jin and Juan J. de Pablo},
journal= {arXiv preprint arXiv:2504.13575},
year = {2025}
}