Learning residue level protein dynamics with multiscale Gaussians

2025-08-31

AI Summary

Protein dynamics are essential for function, yet molecular dynamics (MD) simulations remain computationally prohibitive for large-scale applications. To address this, we propose DynaProt—a lightweight, SE(3)-invariant neural network framework that directly predicts residue-level dynamic descriptors from static protein structures. Its core innovation lies in jointly modeling anisotropic flexibility and inter-residue covariance via multivariate Gaussian distributions, enabled by multi-scale geometric feature extraction and SE(3)-equivariant message passing for parameter-efficient learning. DynaProt achieves state-of-the-art accuracy in root-mean-square fluctuation (RMSF) prediction, reconstructs full residue-wise covariance matrices, operates 10⁴–10⁵× faster than MD simulations at inference, and reduces model parameters by over 99%. As the first scalable, high-fidelity dynamic surrogate model for structural biology, DynaProt enables efficient, physics-informed analysis of protein conformational ensembles without costly simulations.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Search and Optimization: Sampling/Simulation-based Search
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Many methods have been developed to predict static protein structures, however understanding the dynamics of protein structure is essential for elucidating biological function. While molecular dynamics (MD) simulations remain the in silico gold standard, its high computational cost limits scalability. We present DynaProt, a lightweight, SE(3)-invariant framework that predicts rich descriptors of protein dynamics directly from static structures. By casting the problem through the lens of multivariate Gaussians, DynaProt estimates dynamics at two complementary scales: (1) per-residue marginal anisotropy as $3 imes 3$ covariance matrices capturing local flexibility, and (2) joint scalar covariances encoding pairwise dynamic coupling across residues. From these dynamics outputs, DynaProt achieves high accuracy in predicting residue-level flexibility (RMSF) and, remarkably, enables reasonable reconstruction of the full covariance matrix for fast ensemble generation. Notably, it does so using orders of magnitude fewer parameters than prior methods. Our results highlight the potential of direct protein dynamics prediction as a scalable alternative to existing methods.

Problem

• Predicts protein dynamics from static structures efficiently
• Estimates residue-level flexibility and pairwise dynamic couplings
• Provides scalable alternative to molecular dynamics simulations

Innovation

• SE(3)-invariant framework predicts protein dynamics
• Uses multivariate Gaussians at two complementary scales
• Lightweight method with orders fewer parameters