Solvation Free Energies from Neural Thermodynamic Integration

2024-10-21

Journal of Chemical Physics

AI Summary

This work addresses the challenge of simultaneously modeling Lennard-Jones (LJ) and electrostatic interactions with high accuracy while accounting for molecular rigid-body rotation in solvation free energy calculations. We propose a novel thermodynamic integration (TI) method based on a neural network potential (NNP) employing a three-body representation. Crucially, Hamiltonian interpolation is defined at the sample-distribution level, ensuring that the NNP strictly satisfies equilibrium physical constraints at every TI step. Our approach achieves, for the first time, joint continuous coupling of LJ and electrostatic terms and explicitly incorporates molecular rigid-body rotation. Integrated with molecular dynamics sampling and high-precision TI integration, the method significantly improves free energy prediction accuracy on atomic-scale benchmarks—including LJ fluids and aqueous systems (e.g., methane insertion into water). This establishes a new paradigm for efficient, high-fidelity simulation of complex solvation processes.

Technology Category

• Machine Learning: Kernel Methods
• Natural Language Processing: Safety and Robustness
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

We present a method for computing free-energy differences using thermodynamic integration with a neural network potential that interpolates between two target Hamiltonians. The interpolation is defined at the sample distribution level, and the neural network potential is optimized to match the corresponding equilibrium potential at every intermediate time step. Once the interpolating potentials and samples are well-aligned, the free-energy difference can be estimated using (neural) thermodynamic integration. To target molecular systems, we simultaneously couple Lennard-Jones and electrostatic interactions and model the rigid-body rotation of molecules. We report accurate results for several benchmark systems: a Lennard-Jones particle in a Lennard-Jones fluid, as well as the insertion of both water and methane solutes in a water solvent at atomistic resolution using a simple three-body neural-network potential.

Problem

• Computing solvation free energies accurately
• Interpolating Hamiltonians with neural networks
• Modeling molecular interactions and rotations

Innovation

• Neural network potential for thermodynamic integration
• Interpolation at sample distribution level
• Modeling rigid-body rotation and interactions