Scalable emulation of protein equilibrium ensembles with generative deep learning
Масштабируемое моделирование равновесных ансамблей белков с помощью генеративного глубокого обучения
2025-07-10
SCID: 54.1/4f2jmms6
Discuss with AI
BioEmugenerative deep learningmolecular dynamics simulationsprotein equilibrium ensemblesrelative free energies
Figures from the paper
Abstract (AI)
Following the sequence and structure revolutions, predicting functionally relevant protein structure changes at scale remains an outstanding challenge. We introduce BioEmu, a deep learning system that emulates protein equilibrium ensembles by generating thousands of statistically independent structures per hour on a single graphics processing unit (GPU). BioEmu integrates more than 200 milliseconds of molecular dynamics (MD) simulations, static structures, and experimental protein stabilities using new training algorithms. It captures diverse functional motions-including cryptic pocket formation, local unfolding, and domain rearrangements-and predicts relative free energies with 1 kilocalorie per mole accuracy compared with millisecond-scale MD and experimental data. BioEmu provides mechanistic insights by jointly modeling structural ensembles and thermodynamic properties. This approach amortizes the cost of MD and experimental data generation, demonstrating a scalable path toward understanding and designing protein function.
Key Findings
1
BioEmu captures diverse functionally relevant motions, including cryptic pocket formation, local unfolding, and domain rearrangements.
2
BioEmu is a deep learning system that emulates protein equilibrium ensembles, generating thousands of statistically independent structures per hour on a single GPU.
3
It predicts relative free energies with 1 kilocalorie per mole accuracy compared with millisecond-scale molecular dynamics and experimental data.
4
Joint modeling of structural ensembles and thermodynamic properties provides mechanistic insights while amortizing the cost of molecular dynamics and experimental data generation.
5
The model integrates more than 200 milliseconds of molecular dynamics simulations, static structures, and experimental protein stabilities using new training algorithms.
Research Object
Protein equilibrium ensembles and their functionally relevant structural changes
Research Subject
Scalable generation and characterization of protein conformational diversity, functional motions, and relative free energies
Publication Details
Publication Date
2025-07-10
Journal
Publisher
ISSN
Cited by
291
Access Type
Author Information
Download PDF
Subscribe to digest
References available in scid.ai6
Highly accurate protein structure prediction with AlphaFold2021
Comparison of simple potential functions for simulating liquid water1983
GROMACS: High performance molecular simulations through multi-level parallelism from laptops to supercomputers2015
Accurate structure prediction of biomolecular interactions with AlphaFold 32024
Accurate prediction of protein structures and interactions using a three-track neural network2021
A survey of uncertainty in deep neural networks2023