Real space in cryo-EM: the future is local.

Area of Science:

• Structural Biology
• Biophysics
• Computational Biology

Background:

• Cryo-electron microscopy (cryo-EM) yields low signal-to-noise ratio (SNR) images due to radiation sensitivity and poor contrast of biological macromolecules.
• Current cryo-EM image processing relies on averaging, parameter optimization, and prior information, but struggles with parameter estimation, overfitting, and signal variations.
• Traditional Fourier-space methods have limitations in handling local variations inherent in biological samples.

Purpose of the Study:

• To propose and evaluate novel image processing strategies for cryo-EM that better address the local nature of biological samples.
• To improve the reliability and accuracy of macromolecular structure determination from low-SNR cryo-EM data.
• To explore the integration of real-space and Fourier-space approaches for enhanced cryo-EM image processing.

Main Methods:

• Development of real-space measures and filters that account for local variations within cryo-EM images and volumes.
• Application of band-pass filtered real-space volumes and striding resolution through Fourier space.
• Exploration of incorporating powerful prior information, such as from AlphaFold, within real-space processing frameworks.

Main Results:

• Real-space measures are demonstrated to be more reliable and appropriate for biological samples than global Fourier-space measures.
• Proposed hybrid real-space/Fourier-space methods are expected to outperform global Fourier-space-based approaches.
• Locality in image processing, through real-space operations, is identified as a central theme for future cryo-EM advancements.

Conclusions:

• Real-space processing offers significant advantages for cryo-EM by addressing local sample variations and enabling novel prior information integration.
• A combination of real-space operations on frequency bands and striding resolution offers a promising direction for improved cryo-EM image processing.
• Future cryo-EM image processing will likely integrate real-space locality with advanced computational methods, including deep learning and structural prediction tools.