Align Your Structures: Generating Trajectories with Structure Pretraining for Molecular Dynamics

2026-04-04

AI Summary

This work addresses the challenges of data scarcity and high-dimensional distribution modeling in molecular dynamics trajectory generation by proposing a novel approach that integrates structural pretraining with trajectory interpolation. The method first pretrains a diffusion model on large-scale static conformational data to capture molecular geometric priors, then introduces an interpolation module to enforce temporal consistency, decomposing trajectory generation into structure synthesis and temporal alignment as two subtasks. By uniquely combining structural pretraining with dynamic interpolation, the framework effectively leverages abundant static data to mitigate the paucity of trajectory data. Experiments demonstrate that the generated trajectories significantly outperform existing methods across QM9, DRUGS, and peptide/protein systems in terms of geometric plausibility, dynamical consistency, and energetic accuracy, exhibiting enhanced chemical realism.

Abstract

Generating molecular dynamics (MD) trajectories using deep generative models has attracted increasing attention, yet remains inherently challenging due to the limited availability of MD data and the complexities involved in modeling high-dimensional MD distributions. To overcome these challenges, we propose a novel framework that leverages structure pretraining for MD trajectory generation. Specifically, we first train a diffusion-based structure generation model on a large-scale conformer dataset, on top of which we introduce an interpolator module trained on MD trajectory data, designed to enforce temporal consistency among generated structures. Our approach effectively harnesses abundant structural data to mitigate the scarcity of MD trajectory data and effectively decomposes the intricate MD modeling task into two manageable subproblems: structural generation and temporal alignment. We comprehensively evaluate our method on the QM9 and DRUGS small-molecule datasets across unconditional generation, forward simulation, and interpolation tasks, and further extend our framework and analysis to tetrapeptide and protein monomer systems. Experimental results confirm that our approach excels in generating chemically realistic MD trajectories, as evidenced by remarkable improvements of accuracy in geometric, dynamical, and energetic measurements.

Problem

• molecular dynamics
• trajectory generation
• data scarcity
• high-dimensional distribution
• structural consistency

Innovation

• structure pretraining
• diffusion model
• molecular dynamics trajectory
• temporal consistency
• conformer generation