Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Protein Graph Neural Networks for Heterogeneous Cryo-EM Reconstruction

Jonathan Krook, Axel Janson, Joakim andén, Melanie Weber, Ozan Öktem

TL;DR

A geometry-aware method for heterogeneous single-particle cryogenic electron microscopy (cryo-EM) reconstruction that predicts atomic backbone conformations and uses a graph neural network autodecoder to incorporate protein-structure priors, highlighting the benefits of a geometry-informed inductive bias.

Abstract

We present a geometry-aware method for heterogeneous single-particle cryogenic electron microscopy (cryo-EM) reconstruction that predicts atomic backbone conformations. To incorporate protein-structure priors, we represent the backbone as a graph and use a graph neural network (GNN) autodecoder that maps per-image latent variables to 3D displacements of a template conformation. The objective combines a data-discrepancy term based on a differentiable cryo-EM forward model with geometric regularization, and it supports unknown orientations via ellipsoidal support lifting (ESL) pose estimation. On synthetic datasets derived from molecular dynamics trajectories, the proposed GNN achieves higher accuracy compared to a multilayer perceptron (MLP) of comparable size, highlighting the benefits of a geometry-informed inductive bias.

Protein Graph Neural Networks for Heterogeneous Cryo-EM Reconstruction

TL;DR

A geometry-aware method for heterogeneous single-particle cryogenic electron microscopy (cryo-EM) reconstruction that predicts atomic backbone conformations and uses a graph neural network autodecoder to incorporate protein-structure priors, highlighting the benefits of a geometry-informed inductive bias.

Abstract

We present a geometry-aware method for heterogeneous single-particle cryogenic electron microscopy (cryo-EM) reconstruction that predicts atomic backbone conformations. To incorporate protein-structure priors, we represent the backbone as a graph and use a graph neural network (GNN) autodecoder that maps per-image latent variables to 3D displacements of a template conformation. The objective combines a data-discrepancy term based on a differentiable cryo-EM forward model with geometric regularization, and it supports unknown orientations via ellipsoidal support lifting (ESL) pose estimation. On synthetic datasets derived from molecular dynamics trajectories, the proposed GNN achieves higher accuracy compared to a multilayer perceptron (MLP) of comparable size, highlighting the benefits of a geometry-informed inductive bias.
Paper Structure (12 sections, 5 equations, 4 figures, 1 table, 1 algorithm)

This paper contains 12 sections, 5 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Problem Formulation
  3. Method
  4. Forward Model
  5. Pose Estimation
  6. GNN Autodecoder Model
  7. Optimization
  8. Experimental Validation
  9. Metrics
  10. Datasets
  11. Results
  12. Conclusion

Figures (4)

  • Figure 1: Principal overview of our method. A data point in the form of an EM image is indexed by a low dimensional variable, and then mapped via a GNN to a reconstructed conformation. Blue indicates components that are only used during optimization, where the deformed template is posed and compared to data, and regularization is added.
  • Figure 2: Top: Closed and open conformations of ADK. Bottom: Heterogeneous conformations of NSP.
  • Figure 3: Histogram of ground truth RMSD values for all 102k conformations in the ADK dataset, before and after optimizing the GNN autodecoder with $\mathcal{R}_2$-regularization and ESL.
  • Figure 4: Representative reconstruction. (a): Ground truth conformation. (b): Simulated cryo-EM image with randomized orientation. (c): Template conformation generated by AlphaFold 3, with a ground truth RMSD of 6.93 Å. (d): Final prediction by GNN autodecoder, with a ground truth RMSD of 1.85 Å.