# Align Your Structures: Generating Trajectories with Structure Pretraining for Molecular Dynamics Source: [Align Your Structures: Generating Trajectories with Structure Pretraining for Molecular Dynamics]() Collection index: [llms.txt]() Data updated: 2026-04-08T05:00:04.658Z Authors: Aniketh Iyengar, Jiaqi Han, Pengwei Sun, Mingjian Jiang, Jianwen Xie, Stefano Ermon Published: 2026-04-05T00:37:28.000Z ## 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. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Molecular Energy Prediction on Drugs (test)](): Molecular energy prediction; Drugs (test); Energy (Hartrees): 959.94. - [Energy Profile Prediction on GEOM-Drugs (test)](): Energy profile prediction; GEOM-DRUGS (test); W1 Distance: 0.2454. - [Conformer Generation on GEOM-Drugs (test)](): Conformer Generation; GEOM-DRUGS (test); Mean Torsional Decorrelation Time (ps): 5.2. - [Energy Profile Prediction on QM9 (test)](): Energy profile prediction; QM9 (test); W1 Distance: 0.8127. - [Forward Simulation on Drugs](): Forward Simulation; Drugs; JSD Mean Bond Angle Deviation: 0.173. - [Unconditional Generation on QM9 (Geometric Properties)](): Unconditional Generation; QM9; JSD Mean Bond Angle: 0.305. - [Conformer Generation on QM9 (test)](): Conformer Generation; QM9 (test); Mean Torsional Decorrelation Time (ps): 13.59. - [Energy profile prediction on Tetrapeptides (test)](): Energy profile prediction; Tetrapeptides (test); W1: 0.3806. - [Molecular Dynamics Trajectory Generation on MDGen Tetrapeptide](): Molecular Dynamics Trajectory Generation; MDGen Tetrapeptide; Torsion BB: 0.378. - [Molecular Energy Prediction on QM9 (test)](): Molecular energy prediction; QM9 (test); Energy (Hartrees): 388.225. ## Original sources and implementation [Original paper]() [PDF]()