3D Flexible Refinement

Structure and Motion of Flexible Proteins from Cryo-EM

We plan to make an experimental version of 3DFlex available in cryoSPARC. In the meantime, if you have an interesting dataset and would like to collaborate on uncovering the motion of a flexible protein mechanism, please reach out to us!

Single particle cryo-EM excels in determining static structures of biological macromolecules such as proteins. However, many proteins are dynamic, with their motion inherently linked to their function. Recovering the continuous motion and detailed 3D structure of flexible proteins from cryo-EM data has remained an open challenge.

We introduce 3D Flexible Refinement (3DFlex), a motion-based deep neural network model of continuous heterogeneity. 3DFlex directly exploits the knowledge that conformational variability of a protein is often the result of physical processes that transport density over space and tend to conserve mass and preserve local geometry.

From 2D image data, the 3DFlex model jointly learns a single canonical 3D map, latent coordinate vectors that specify positions on the protein’s conformational landscape, and a flow generator that, given a latent position as input, outputs a 3D deformation field. This deformation field convects the canonical map into appropriate conformations to explain experimental images.

Applied to experimental data, 3DFlex learns non-rigid motion spanning several orders of magnitude while preserving high-resolution details of secondary structure elements. Further, 3DFlex resolves canonical maps that are improved relative to conventional refinement methods because particle images contribute to the maps coherently regardless of the conformation of the protein in the image.

Together, the ability to obtain insight into motion in macromolecules, as well as the ability to resolve features that are usually lost in cryo-EM of flexible specimens, will provide new insight and allow new avenues of investigation into biomolecular structure and function.

3DFlex Model
Figure 1. The 3DFlex model represents the flexible 3D structure of a protein as deformations of a single canonical 3D density map V. Under the model, a single particle image is associated with a low-dimensional latent coordinate z that encodes the conformation for the particle in the image. A neural flow generator network fθ converts the latent coordinate into the flow field u and a convection operator then deforms the canonical density to generate a convected map W. This map can then be projected along the particle viewing direction determined by the pose φ, CTF-corrupted, and compared against the experimental image. In 3DFlex, the canonical 3D map, the latent coordinates, and the flow generator are all learned from experimental data, allowing the method to reveal non-rigid motion and improved structural features in flexible regions.

Videos of protein motion

The flexible motion learned by 3DFlex is difficult to visualize in a static image. The videos below show more clearly the detailed motion that is revealed and the quality of the reconstruction that is enabled by virtue of flexible refinement.

Supplementary Video 1. This video shows results of 3DFlex on a dataset of 102,500 particle images of a tri-snRNP spliceosome particle (EMPIAR-10073). 3DFlex is run with a K=5-dimensional latent space, and different regions of the space correspond to different parts of the particle's conformational landscape. This video shows the output of the 3DFlex generative model as latent coordinates are varied along three axes (coordinates 1, 3, and 5). These dimensions encode non-rigid motion of the head region of the protein, where different parts and subunits move and bend relative to eachother.
Supplementary Video 2. This video shows results of 3DFlex on a dataset of 200,00 particle images of a TRPV1 ion channel (EMPIAR-10059). 3DFlex is run with a K=2-dimensional latent space. The video shows the output of the 3DFlex generative model as latent coordinates are varied along each of the two dimensions. The first dimension reveals inward and outward coordinated bending of opposite flexible subunits in the soluble domain. The second dimension reveals twisting of the subunits around the pore axis.
Supplementary Video 3. This video shows a comparison between the reconstructed density map from a conventional refinement and flexible refinement using 3DFlex for the TRPV1 ion channel. Map quality and local resolutions are substantially improved in the peripheral helices. Notably, local focused refinement using a mask around the flexible part cannot improve the reconstruction compared to a conventional refinement, because the flexible parts are non-rigid and too small for individual pose alignment.