Subspace Method of Moments for Ab Initio 3-D Single Particle Cryo-EM Reconstruction

Area of Science:

• Structural Biology
• Computational Imaging
• Biophysics

Background:

• Cryo-electron microscopy (cryo-EM) reconstructs 3-D molecular structures from noisy 2-D projection images.
• Accurate reconstruction is hindered by unknown viewing angles and low signal-to-noise ratios.
• Traditional methods often require computationally intensive angle estimation.

Purpose of the Study:

• To develop a computationally efficient method for cryo-EM structure reconstruction.
• To address the challenges of noise and unknown viewing angles in cryo-EM data.
• To improve the resolution of reconstructed 3-D molecular structures.

Main Methods:

• Proposed the subspace method of moments (SubspaceMoM) to compress image moments.
• Utilized data-driven low-rank tensor techniques and function basis expansion for compression.
• Employed numerical quadrature for efficient computation and joint reconstruction of structure and viewing angle distribution.

Main Results:

• SubspaceMoM significantly improves reconstruction resolution compared to previous moment-based methods.
• The method demonstrates practical applicability on synthetic datasets with a simplified cryo-EM model.
• Achieved better resolution using up to the third-order moment.

Conclusions:

• SubspaceMoM offers a computationally efficient and robust alternative for cryo-EM structure determination.
• The technique effectively mitigates computational and memory costs associated with higher-order moments.
• This advancement holds promise for more accurate and accessible structural biology research.