A Bayesian approach to beam-induced motion correction in cryo-EM single-particle analysis

Area of Science:

• Structural Biology
• Biophysics
• Microscopy

Background:

• Accurate particle motion estimation and radiation damage assessment are critical for high-resolution cryo-EM.
• Existing methods may lack robustness in modeling complex motion and damage effects.

Purpose of the Study:

• To develop and validate a novel computational method for estimating particle trajectories and cumulative beam damage in cryo-EM single-particle analysis.
• To improve the accuracy and robustness of structural determination by accounting for sample motion and radiation effects.

Main Methods:

• Utilized Gaussian process regression to model sample motion, incorporating a prior favoring spatial and temporal smoothness.
• Developed an efficient stochastic method to estimate parameters for the smoothness prior.
• Implemented a practical algorithm to estimate cumulative radiation damage and fit relative B-factors.

Main Results:

• The new Bayesian polishing method was evaluated on three public datasets.
• Demonstrated superior performance compared to state-of-the-art methods and previous approaches.
• Successfully integrated into RELION-3, replacing the existing particle polishing method.

Conclusions:

• The proposed Bayesian polishing method offers a significant advancement in cryo-EM data processing.
• Accurate modeling of particle motion and radiation damage leads to improved structural resolution.
• The method's implementation in RELION-3 provides a practical tool for the cryo-EM community.