Petimot: A Novel Framework for Inferring Protein Motions from Sparse Data Using SE(3)-Equivariant Graph Neural Networks
Abstract
Proteins move and deform to ensure their biological functions. Despite significant progress in protein structure prediction, approximating conformational ensembles under physiological conditions remains a fundamental open problem. This paper presents a novel perspective on the problem by directly targeting continuous compact representations of protein motions inferred from sparse experimental observations. We develop a task-specific loss function enforcing data symmetries, including scaling and permutation operations. Our method PETIMOT (Protein sEquence and sTructure-based Inference of MOTions) leverages transfer learning from pre-trained protein language models through an SE(3)-equivariant graph neural network. When trained and evaluated on the Protein Data Bank, PETIMOT shows superior performance in time and accuracy, capturing protein dynamics, particularly large/slow conformational changes, compared with state-of-the-art diffusion and flow-matching approaches, as well as traditional physics-based models. Our code and protocols are available at https://github.com/PhyloSofS-Team/PETIMOT.
Related
- Shaping Slow Manifolds in Equivariant Recurrent Neural Networks
- Maximum likelihood, neural networks and posterior mean for inferring fractional Brownian motion parameters
- CohAlign: Predicting the first-order noise susceptibility of gradients in equivariant quantum neural networks
- From wings to tails: a general framework for inferring dispersal potential from morphology
- B-GRASP: A Bayesian Framework for Inferring Graph Weights from SPDE-Inspired Dynamics
BibTeX
@article{lombard2026petimot,
title = {{Petimot: A Novel Framework for Inferring Protein Motions from Sparse Data Using SE(3)-Equivariant Graph Neural Networks}},
author = {Valentin Lombard and Julien Nguyen Van and Sergei Grudinin and Elodie Laine},
journal = {Acta crystallographica Section D : Structural biology [1993-...]},
year = {2026},
month = jul,
doi = {10.1107/s2059798326006054},
url = {https://hal.sorbonne-universite.fr/hal-05334844v2},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Petimot: A Novel Framework for Inferring Protein Motions from Sparse Data Using SE(3)-Equivariant Graph Neural Networks
AU - Valentin Lombard
AU - Julien Nguyen Van
AU - Sergei Grudinin
AU - Elodie Laine
PY - 2026
DA - 2026/07/20
JO - Acta crystallographica Section D : Structural biology [1993-...]
DO - 10.1107/s2059798326006054
UR - https://hal.sorbonne-universite.fr/hal-05334844v2
AB - Proteins move and deform to ensure their biological functions. Despite significant progress in protein structure prediction, approximating conformational ensembles under physiological conditions remains a fundamental open problem. This paper presents a novel perspective on the problem by directly targeting continuous compact representations of protein motions inferred from sparse experimental observations. We develop a task-specific loss function enforcing data symmetries, including scaling and permutation operations. Our method PETIMOT (Protein sEquence and sTructure-based Inference of MOTions) leverages transfer learning from pre-trained protein language models through an SE(3)-equivariant graph neural network. When trained and evaluated on the Protein Data Bank, PETIMOT shows superior performance in time and accuracy, capturing protein dynamics, particularly large/slow conformational changes, compared with state-of-the-art diffusion and flow-matching approaches, as well as traditional physics-based models. Our code and protocols are available at https://github.com/PhyloSofS-Team/PETIMOT.
SN - 2059-7983
SN - 2059-7983
ER -