On the design space between molecular mechanics and machine learning force fields

Abstract

A force field as accurate as quantum mechanics (QM) and as fast as molecular mechanics (MM), with which one can simulate a biomolecular system efficiently enough and meaningfully enough to get quantitative insights, is among the most ardent dreams of biophysicists -- a dream, nevertheless, not to be fulfilled any time soon. Machine learning force fields (MLFFs) represent a meaningful endeavor towards this direction, where differentiable neural functions are parametrized to fit ab initio energies, and furthermore forces through automatic differentiation. We argue that, as of now, the utility of the MLFF models is no longer bottlenecked by accuracy but primarily by their speed (as well as stability and generalizability), as many recent variants, on limited chemical spaces, have long surpassed the chemical accuracy of $1$ kcal/mol -- the empirical threshold beyond which realistic chemical predictions are possible -- though still magnitudes slower than MM. Hoping to kindle explorations and designs of faster, albeit perhaps slightly less accurate MLFFs, in this review, we focus our attention on the design space (the speed-accuracy tradeoff) between MM and ML force fields. After a brief review of the building blocks of force fields of either kind, we discuss the desired properties and challenges now faced by the force field development community, survey the efforts to make MM force fields more accurate and ML force fields faster, envision what the next generation of MLFF might look like.

References in corpus (96)

• Array Programming with NumPy
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Semi-Supervised Classification with Graph Convolutional Networks
• Denoising Diffusion Probabilistic Models
• Neural Message Passing for Quantum Chemistry
• SchNet - a deep learning architecture for molecules and materials
• On the Opportunities and Risks of Foundation Models
• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Diffusion Models in Vision: A Survey
• Building Water Models, A Different Approach
• Score-Based Generative Modeling through Stochastic Differential Equations
• Machine Learning of Accurate Energy-Conserving Molecular Force Fields
• Simplifying Graph Convolutional Networks
• PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments and Partial Charges
• Machine learning for molecular simulation
• Convolutional Networks on Graphs for Learning Molecular Fingerprints
• Linformer: Self-Attention with Linear Complexity
• Less is more: sampling chemical space with active learning
• The Open Catalyst 2020 (OC20) Dataset and Community Challenges
• Deep Graph Library: A Graph-Centric, Highly-Performant Package for Graph Neural Networks
• Computer Modeling of Halogen Bonds and Other $\sigma$-Hole Interactions
• Neural Ordinary Differential Equations
• Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
• A Fourth-Generation High-Dimensional Neural Network Potential with Accurate Electrostatics Including Non-local Charge Transfer
• SchNet: A continuous-filter convolutional neural network for modeling quantum interactions
• How Powerful are Graph Neural Networks?
• SpookyNet: Learning Force Fields with Electronic Degrees of Freedom and Nonlocal Effects
• ANI-1: A data set of 20M off-equilibrium DFT calculations for organic molecules
• Biological and synthetic membranes: What can be learned from a coarse-grained description?
• MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields
• Folding@Home and Genome@Home: Using distributed computing to tackle previously intractable problems in computational biology
• Equivariant message passing for the prediction of tensorial properties and molecular spectra
• OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features
• Principal Neighbourhood Aggregation for Graph Nets
• TorchMD: A deep learning framework for molecular simulations
• Topology Adaptive Graph Convolutional Networks
• MoFlow: An Invertible Flow Model for Generating Molecular Graphs
• GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation
• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• Coarse-Graining Auto-Encoders for Molecular Dynamics
• Forces are not Enough: Benchmark and Critical Evaluation for Machine Learning Force Fields with Molecular Simulations
• Machine Learning in QM/MM Molecular Dynamics Simulations of Condensed-Phase Systems
• Diffusion Models: A Comprehensive Survey of Methods and Applications
• QM7-X: A comprehensive dataset of quantum-mechanical properties spanning the chemical space of small organic molecules
• Molecular geometry prediction using a deep generative graph neural network
• Equivariant Diffusion for Molecule Generation in 3D
• The Design Space of E(3)-Equivariant Atom-Centered Interatomic Potentials
• End-to-End Differentiable Molecular Mechanics Force Field Construction
• Open Materials 2024 (OMat24) Inorganic Materials Dataset and Models
• A Note on Over-Smoothing for Graph Neural Networks
• On the Bottleneck of Graph Neural Networks and its Practical Implications
• Ultra-fast interpretable machine-learning potentials
• A Survey on Oversmoothing in Graph Neural Networks
• Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks
• Machine Learning Directed Optimization of Classical Molecular Modeling Force Fields
• Stability of Complex Biomolecular Structures: Vander Waals, Hydrogen Bond Cooperativity, and Nuclear Quantum Effects
• Equivariant Flows: Exact Likelihood Generative Learning for Symmetric Densities
• MACE-OFF: Transferable Short Range Machine Learning Force Fields for Organic Molecules
• TorchMD-NET: Equivariant Transformers for Neural Network based Molecular Potentials
• Generalization and Representational Limits of Graph Neural Networks
• BuRNN: Buffer Region Neural Network Approach for Polarizable-Embedding Neural Network / Molecular Mechanics Simulations
• Machine-learned molecular mechanics force field for the simulation of protein-ligand systems and beyond
• Scalable Hybrid Deep Neural Networks/Polarizable Potentials Biomolecular Simulations including long-range effects
• Force-Field-Enhanced Neural Network Interactions: from Local Equivariant Embedding to Atom-in-Molecule properties and long-range effects
• Multiple Time Step Integrators in Ab Initio Molecular Dynamics
• Reactive Atomistic Simulations of Diels-Alder Reactions: the Importance of Molecular Rotations
• Understanding over-squashing and bottlenecks on graphs via curvature
• Differentiable Molecular Simulations for Control and Learning
• Scalars are universal: Equivariant machine learning, structured like classical physics
• Elucidating the proton transport pathways in liquid imidazole with first-principles molecular dynamics
• Grappa -- A Machine Learned Molecular Mechanics Force Field
• Reducing SO(3) Convolutions to SO(2) for Efficient Equivariant GNNs
• EspalomaCharge: Machine learning-enabled ultra-fast partial charge assignment
• On the Expressive Power of Geometric Graph Neural Networks
• A $\Delta$-Machine Learning Approach for Force Fields, Illustrated by a CCSD(T) 4-body Correction to the MB-pol Water Potential
• Timewarp: Transferable Acceleration of Molecular Dynamics by Learning Time-Coarsened Dynamics
• Equivariant Graph Mechanics Networks with Constraints
• TensorNet: Cartesian Tensor Representations for Efficient Learning of Molecular Potentials
• Graph Nets for Partial Charge Prediction
• DASH: Dynamic Attention-Based Substructure Hierarchy for Partial Charge Assignment
• Scaling the leading accuracy of deep equivariant models to biomolecular simulations of realistic size
• Chemically Transferable Generative Backmapping of Coarse-Grained Proteins
• Spatial Attention Kinetic Networks with E(n)-Equivariance
• Frame Averaging for Invariant and Equivariant Network Design
• Stochastic Aggregation in Graph Neural Networks
• Transferable Boltzmann Generators
• FAENet: Frame Averaging Equivariant GNN for Materials Modeling
• Molecular relaxation by reverse diffusion with time step prediction
• Non-convolutional Graph Neural Networks
• Machine Learning Potentials: A Roadmap Toward Next-Generation Biomolecular Simulations
• Enabling Efficient Equivariant Operations in the Fourier Basis via Gaunt Tensor Products
• SE(3) Equivariant Augmented Coupling Flows
• Equivariance Is Not All You Need: Characterizing the Utility of Equivariant Graph Neural Networks for Particle Physics Tasks
• "Hey, that's not an ODE": Faster ODE Adjoints via Seminorms

Cited by in corpus (4)

• NepoIP/MM: Towards Accurate Biomolecular Simulation with a Machine Learning/Molecular Mechanics Model Incorporating Polarization Effects
• Machine Learning Enhanced Calculation of Quantum-Classical Binding Free Energies
• Efficient GPU-Accelerated Training of a Neuroevolution Potential with Analytical Gradients
• Machine-Learned Potentials for Solvation Modeling