Practical and Accurate Free Energy Calculations using Neural Network Potentials

This poster presents our approach to combining neural network potentials with traditional free energy methods to achieve quantum-level accuracy at a fraction of the computational cost.

Key Highlights

• Neural network potentials trained on DFT data
• 10-100x speedup compared to traditional QM/MM methods
• Validated on diverse protein-ligand systems
• Integration with Deep Origin’s simulation platform

Applications

Our method enables practical free energy calculations for:

• Relative binding free energy predictions
• Solvation free energy calculations
• pKa predictions