Accelerating Protein Molecular Dynamics Simulation with DeepJump

2025-09-16

AI Summary

Traditional molecular dynamics (MD) simulations are computationally prohibitive for accessing biologically relevant timescales of protein conformational dynamics. To address this, we propose DeepJump—a novel deep generative model that integrates Euclidean-equivariant neural networks with flow matching, enabling generalizable modeling of protein conformational dynamics across multiple timescales. Trained on the mdCATH protein trajectory dataset, DeepJump supports both *de novo* folding pathway prediction and native-state sampling. Experiments demonstrate that DeepJump achieves approximately 1000× speedup over conventional MD while preserving dynamical fidelity: it accurately reproduces long-timescale conformational evolution and successfully simulates *de novo* folding. Moreover, our analysis reveals fundamental trade-offs between computational efficiency and dynamical accuracy, establishing a new paradigm for efficient, interpretable, deep learning–based protein dynamics modeling.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Unraveling the dynamical motions of biomolecules is essential for bridging their structure and function, yet it remains a major computational challenge. Molecular dynamics (MD) simulation provides a detailed depiction of biomolecular motion, but its high-resolution temporal evolution comes at significant computational cost, limiting its applicability to timescales of biological relevance. Deep learning approaches have emerged as promising solutions to overcome these computational limitations by learning to predict long-timescale dynamics. However, generalizable kinetics models for proteins remain largely unexplored, and the fundamental limits of achievable acceleration while preserving dynamical accuracy are poorly understood. In this work, we fill this gap with DeepJump, an Euclidean-Equivariant Flow Matching-based model for predicting protein conformational dynamics across multiple temporal scales. We train DeepJump on trajectories of the diverse proteins of mdCATH, systematically studying our model's performance in generalizing to long-term dynamics of fast-folding proteins and characterizing the trade-off between computational acceleration and prediction accuracy. We demonstrate the application of DeepJump to ab initio folding, showcasing prediction of folding pathways and native states. Our results demonstrate that DeepJump achieves significant $approx$1000$ imes$ computational acceleration while effectively recovering long-timescale dynamics, providing a stepping stone for enabling routine simulation of proteins.

Problem

• Accelerating protein molecular dynamics simulations computationally
• Predicting long-timescale protein conformational dynamics accurately
• Generalizing kinetics models across diverse protein systems

Innovation

• Euclidean-Equivariant Flow Matching model
• Predicts protein conformational dynamics
• Achieves 1000x computational acceleration