Overview

An increasing volume of experimental and molecular dynamics (MD) simulation data has become available for biomolecular systems. However, a persistent gap has long existed between low-resolution but highly reliable experimental measurements and the high-resolution but computationally expensive, short-timescale MD simulations. This limitation has hindered comprehensive characterization of biomolecular systems, their (mal)functions, and the development of therapeutic interventions.

To address this challenge, we develop a machine-learned coarse-grained (MLCG) model that is capable of routinely reaching biologically relevant timescales while maintaining accuracy comparable to atomistic force fields. The model is designed to be systematically refined using experimental data, enabling improved fidelity across scales. By leveraging modern machine learning methods, we are working to establish such a model and release it to the wider scientific community. Beyond its immediate impact in biophysics, this work may inspire similar advances in adjacent fields, including materials science.
For more information on our research, click on Our Research