Description
Machine learning transforms molecular modelling by learning potential energy surfaces and force fields directly from quantum mechanical data, and these methods are increasingly applied to predict properties and run molecular dynamics simulations. This hands-on workshop bridges theory and practice, introducing the core physics behind ML potentials while you apply them using pre-configured OpenMM environments and pretrained models including ANI2X, AIMNet2, MACE-OFF24, OrbMol-v3, and FeNNix-Bio1M. You compare approaches, explore their practical trade-offs, and run hybrid ML/MM simulations that merge machine-learned potentials with classical molecular mechanics. By the end, you grasp the key ideas distinguishing major ML potential families, gain practical experience running ML-driven MD, and learn how different ML potentials suit different molecular systems and research goals, from small molecules to complex biomolecular systems. You leave with ready-to-use workflows and the confidence to integrate these methods into your own research.
Prerequisites
Participants need some familiarity with molecular dynamics or molecular simulation concepts, such as molecular structures, potential energy, forces, and trajectories. Basic Python skills sufficient to read and modify simple Python scripts are also expected. No prior experience with machine-learning potentials or OpenMM is necessary.
Details
Learning Objectives
- Explain the physical principles and key ideas underlying major machine-learning potential approaches.
- Run molecular dynamics simulations using OpenMM and OpenMM-ML with pretrained ML potentials.
- Configure and modify basic Python/OpenMM input scripts for ML-driven MD simulations.
- Use pre-trained ANI2X, AIMNet2, MACE-OFF24, OrbMol-v3, and FeNNix-Bio1M. models in ready-to-use environments.
- Compare ML potentials and assess their suitability for different molecular systems.
- Set up and run hybrid ML/MM molecular dynamics simulations.
Who Should Enrol
Graduate students, postdoctoral researchers, and researchers in chemistry, biochemistry, materials science, computational chemistry, and related fields who are interested in molecular modelling, molecular dynamics, and machine-learning potentials.