SOURCE-LINKED INTELLIGENCE

Protein Dynamics with Generalized machine-learned potentials

CORDIS · observation · Publication date unknown

proach that matches atomistic accuracy while remaining transferable and predictive is still lacking. Building on a recent key milestone in MLCG research, this project will advance a physics-informed machine learning framework that learns many-body CG potentials from high-resolution simulations and experimental data. The resulting model will be chemically transferable, physically interpretable, and capable of predicting conformational dynamics, free energy changes, and binding affinities across diverse biomolecular systems. By integrating statistical physics, graph neural networks, and experimental validation, the project will overcome key limitations of current CG models—such as the treatment of long-range interactions—and demonstrate its power through applications to biomedically relevant systems. Our expertise in MLCG combined with ERC support will enable the focused development and dissemination of a scalable biomolecular modeling framework with broad impact in biophysics and beyond. molecular dynamics, machine learning, proteins, coarse-graining, molecular simulation

Read original source ↗ Open in workspace

recordType

award

status

SIGNED

region

EU

value

3036526

unit

EUR