An Efficient Method to Quantify Structural Distributions in Heterogeneous cryo-EM Datasets

Hanlin Gu, Wei Wang, Ilona Christy Unarta, Wenqi Zeng, Fu Kit Sheong, Peter Pak‐Hang Cheung, Song Liu, Yuan Yao, Xuhui HuangbioRxiv 2021 · DOI

Abstract: Cryogenic Electron Microscopy (cryo-EM) preserves the ensemble of protein conformations in solution, providing a promising way to characterize conformational changes underlying protein functions. However, elucidating distributions of multiple conformations from heterogeneous datasets remains challenging. We developed Linear Combinations of Template Conformations (LCTC), an algorithm that assigns 2D images to template 3D structures via a novel two-stage matching process. By utilizing auto-correlation functions for rapid initial filtering followed by precise pixel-pixel matching, LCTC efficiently quantifies structural populations. We validate LCTC using simulated Taq RNA polymerase (RNAP) datasets, where it outperforms manifold embedding and multi-body refinement, and successfully apply it to experimental E. coli RNAP data.

Introduction

Proteins are dynamic molecular machines that transition between various conformational states to execute biological functions. While cryo-EM can capture snapshots of these ensembles, a significant bottleneck is structural heterogeneity. A typical dataset contains thousands of 2D projections representing different orientations and conformations. While current software excels at high-resolution 3D reconstruction, accurately quantifying the relative populations of these states—essential for understanding thermodynamics and free energy landscapes—remains difficult.

The LCTC Algorithm

The Linear Combinations of Template Conformations (LCTC) method addresses this challenge through a four-step workflow designed for efficiency and accuracy:

  1. Identification of the Best Viewing Angle: LCTC identifies orientations that maximize the variance between template structures using Principal Component Analysis (PCA). This ensures that subtle conformational differences are visible in 2D projections.
  2. Denoising and Projection: Experimental 2D images are denoised, while 3D templates are projected into 2D around the optimal viewing angle.
  3. Two-Stage Matching:
    • Stage 1 (ACF Matching): To reduce computational overhead, LCTC uses rotationally and translationally invariant Auto-correlation Functions (ACF) of image contours to filter out approximately 90% of irrelevant images.
  4. Stage 2 (Pixel-Pixel Matching): The remaining images undergo rigorous pixel-by-pixel intensity comparison for final template assignment.
  5. Quantification: Final populations are determined by the count of images assigned to each structural template.
A schematic overview of the LCTC method, showing the workflow from identifying the best viewing angle to final population estimation.

Results and Discussion

The researchers validated LCTC using a simulated Taq RNAP dataset with three known clamp-opening states and an experimental Eco RNAP dataset containing over 370,000 particles.

The Importance of Viewing Angle

A key insight of this work is the "viewing angle bias." Because conformational changes are often localized, observing a protein from a suboptimal perspective can mask structural transitions. By identifying the "best" angle, LCTC ensures that the 2D data contains sufficient information to distinguish between templates.

Identification of structural differences often depends on the viewing angle; here, a specific orientation is measured to maximize conformational variance.

Performance and Robustness

In simulated tests, LCTC perfectly recovered the ground-truth population distribution (0.4, 0.3, 0.3), whereas traditional clustering-based manifold embedding and maximum likelihood methods showed significant deviations. LCTC also demonstrated remarkable robustness to noise, maintaining accuracy at signal-to-noise ratios (SNR) as low as 0.05. When applied to experimental Eco RNAP data, the algorithm successfully quantified various clamp-opening populations, providing a detailed view of the protein's dynamic equilibrium.

Conclusion

LCTC provides a robust framework for quantifying conformational heterogeneity in cryo-EM. While currently dependent on high-quality 3D templates (often from multi-body refinement), future developments could integrate templates from Molecular Dynamics (MD) simulations. This would enable the quantification of rare or transient states that are otherwise difficult to reconstruct, further bridging the gap between static structures and dynamic biological function.

Summary updated

Abstract

Abstract Cryogenic Electron Microscopy (cryo-EM) preserves the ensemble of protein conformations in solution and thus provide a promising way to characterize conformational changes underlying protein functions. However, it remains challenging for existing software to elucidate distributions of multiple conformations from a heterogeneous cryo-EM dataset. We developed a new algorithm: Linear Combinations of Template Conformations (LCTC) to obtain distributions of multiple conformations from cryo-EM datasets. LCTC assigns 2D images to the template 3D structures obtained by Multi-body Re-finement of RELION via a novel two-stage matching algorithm. Specifically, an initial rapid assignment of experimental 2D images to template 2D images was applied based on auto-correlation functions of image contours that can efficiently remove the majority of irrelevant 2D images. This is followed by pixel-pixel matching of images with fewer number of 2D images, which can accurately assign the 2D images to the template images. We validate the LCTC method by demonstrating that it can accurately reproduce the distributions of 3 Thermus aquaticus ( Taq ) RNA polymerase (RNAP) structures with different degrees of clamp opening from a simulated cryo-EM dataset, in which the correct distributions are known. For this dataset, we also show that LCTC greatly outperforms clustering-based Manifold Embedding and Maximum Likelihood-based Multi-body Re-finement algorithms in terms of reproducing the structural distributions. Lastly, we also successfully applied LCTC to reveal the populations of various clamp-opening conformations from an experimental Escherichia coli RNAP cryo-EM dataset. Source code is available at https://github.com/ghl1995/LCTC .
BibTeX
@misc{gu2021efficient,
  author = {Hanlin Gu and Wei Wang and Ilona Christy Unarta and Wenqi Zeng and Fu Kit Sheong and Peter Pak‐Hang Cheung and Song Liu and Yuan Yao and Xuhui Huang},
  title = {{An Efficient Method to Quantify Structural Distributions in Heterogeneous cryo-EM Datasets}},
  year = {2021},
  doi = {10.1101/2021.05.27.446075},
  howpublished = {bioRxiv},
}
Figure