3DFlex: determining structure and motion of flexible proteins from cryo-EM

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Modeling flexible macromolecules presents a significant challenge in single-particle cryogenic-electron microscopy (cryo-EM).
• Understanding molecular flexibility is crucial for addressing fundamental questions in structural biology.

Purpose of the Study:

• To introduce Three-Dimensional Flexible Refinement (3DFlex), a novel neural network model for analyzing continuous molecular heterogeneity in cryo-EM data.
• To develop a method that models protein motion and conformational landscapes from 2D image data.

Main Methods:

• 3DFlex is a motion-based neural network that leverages physical principles of conformational variability, preserving local geometry.
• The model processes 2D cryo-EM image data to generate high-resolution 3D density maps.
• It explicitly models the motion of flexible proteins across their conformational landscape.

Main Results:

• 3DFlex successfully modeled nonrigid molecular motions for various biological macromolecules, including large complexes and small flexible proteins.
• The method resolved fine details of moving secondary structure elements in tested proteins like the SARS-CoV-2 spike.
• 3DFlex demonstrated improved 3D density resolution compared to existing methods by utilizing coherent signal across the conformational landscape.

Conclusions:

• 3DFlex offers a powerful new approach for modeling molecular flexibility and heterogeneity in cryo-EM.
• The model enhances the resolution and detail obtainable from cryo-EM data, advancing structural biology insights.
• This method provides explicit models of protein dynamics, facilitating a deeper understanding of biological mechanisms.