Glossary

Bayesian Polishing

A per-particle, beam-induced motion correction algorithm implemented in RELION that uses a Bayesian framework to model and reverse radiation damage and movement trajectories.

Get in touch Learn more

Governance lead reviewing model governance framework on laptop, policy documents visible, executive office setup.

CRYO-EM DATA PROCESSING

What is Bayesian Polishing?

A per-particle, beam-induced motion correction algorithm implemented in RELION that uses a Bayesian framework to model and reverse radiation damage and movement trajectories.

Bayesian Polishing is a per-particle motion correction algorithm in RELION that applies a Bayesian framework to model beam-induced specimen movement and radiation damage. It estimates the three-dimensional trajectory of each particle throughout the dose-fractionated movie, using a Gaussian prior on acceleration to regularize the motion tracks and prevent overfitting to noise.

The algorithm operates after an initial 3D reconstruction, re-evaluating the alignment of movie frames against the reference map. By optimally weighting each frame based on its signal-to-noise ratio and the cumulative radiation damage, Bayesian Polishing restores high-resolution detail lost to specimen movement, often yielding significant improvements in map quality and Fourier Shell Correlation (FSC) resolution.

PER-PARTICLE MOTION CORRECTION

Key Features of Bayesian Polishing

Bayesian Polishing is a statistical framework for modeling and reversing beam-induced motion in cryo-EM. It operates on individual particle trajectories, using a prior on motion smoothness to separate true structural signal from radiation damage artifacts.

Bayesian Framework for Motion Trajectories

Models the per-particle motion trajectory as a Gaussian process. The algorithm fits a smooth, continuous path for each particle's movement during the exposure, balancing the fit to the observed data against a prior that penalizes unphysical, jerky motion. This avoids overfitting noise while accurately capturing the true beam-induced drift. The output is a set of corrected particle positions for each movie frame, which are then averaged to produce a sharper reconstruction.

Radiation Damage Compensation

Explicitly accounts for the progressive loss of high-resolution information due to radiation damage. The algorithm applies a per-frame, frequency-dependent weighting that optimally down-weights later frames. This is combined with the motion trajectory model to ensure that the final, averaged particle image extracts the maximum possible signal from early, minimally damaged frames while still using low-resolution information from later frames to improve particle alignment.

Integration with RELION's Refinement Pipeline

Implemented as a native tool within the RELION (REgularised LIkelihood OptimisatioN) software suite. It operates on the output of a standard 3D refinement, using the aligned movie frames and metadata. The polished particle images can then be fed back into a subsequent refinement iteration, often yielding a significant improvement in reported resolution and map interpretability, particularly for flexible or radiation-sensitive specimens.

Comparison to Alternative Motion Correction Methods

Differs fundamentally from global or local patch-based motion correction algorithms like MotionCor2. While those methods fit a single motion model to a whole micrograph or local regions, Bayesian Polishing operates on individual particles. This allows it to capture heterogeneous, particle-specific motion that is averaged out by global approaches, making it especially powerful for samples with significant uncorrelated movement.

Statistical Metrics and Validation

Outputs diagnostic plots showing the fitted trajectories and the B-factor applied to each frame, allowing users to assess the extent of radiation damage and motion. The effectiveness is validated by an increase in the gold-standard Fourier Shell Correlation (FSC) resolution and improved high-resolution features in the final reconstructed density map, confirming that the procedure enhances true signal rather than introducing correlated noise.

BAYESIAN POLISHING EXPLAINED

Frequently Asked Questions

Clarifying the statistical framework behind per-particle motion correction in cryo-EM, addressing common questions about implementation, comparison to other algorithms, and the role of prior probabilities in high-resolution reconstruction.

Bayesian Polishing is a per-particle, beam-induced motion correction algorithm implemented in RELION that uses a Bayesian framework to model and reverse radiation damage and movement trajectories. Unlike global motion correction, it operates on individual particle trajectories extracted from dose-fractionated movie frames. The algorithm constructs a prior probability distribution over possible particle trajectories, favoring smooth, physically plausible paths, and combines this with a likelihood function derived from the fit of the particle images to the current 3D reconstruction. By maximizing the posterior probability, it infers the most likely trajectory for each particle, effectively reversing beam-induced movement and restoring high-resolution signal. The output is a set of polished particles with corrected positions and per-frame dose weights that optimally down-weight radiation-damaged frames, significantly improving map resolvability in the 3-4 Å range.

Enabling Efficiency, Speed & Accuracy

Intelligent Analysis, Decision & Execution

We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.

Talk to Us

Search across company data

Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.

Useful when people spend too long searching or get different answers from different systems.

Enterprise searchRAGPermissions

Automate internal workflows

Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.

Useful when repetitive work moves across multiple tools and teams.

AI agentsWorkflow automationGovernance

Add AI to products and internal tools

Build assistants, guided actions, or decision support into the software your team or customers already use.

Useful when AI needs to be part of the product, not a separate tool.

AI integrationDecision supportModel routing

CRYO-EM MOVIE CORRECTION METHODS

Bayesian Polishing vs. MotionCor2 vs. Dose Weighting

Comparison of computational approaches for correcting beam-induced motion and radiation damage in dose-fractionated cryo-EM movies

Feature	Bayesian Polishing	MotionCor2	Dose Weighting
Primary purpose	Per-particle trajectory estimation and B-factor weighting	Global and local frame alignment	Frequency-dependent SNR weighting per frame
Algorithmic framework	Bayesian inference with Gaussian prior	Patch-based cross-correlation alignment	Exposure-dependent amplitude modulation
Granularity	Per-particle	Per-micrograph (patch-based)	Per-frame (global)
Models radiation damage
Estimates per-particle motion trajectories
Corrects anisotropic motion
Output	Polished particle stack with optimized B-factors	Aligned and averaged micrograph	Weighted frame average
Implementation	RELION	Standalone (GPU-accelerated)	RELION, cryoSPARC, cisTEM
Computational cost	High (per-particle optimization)	Moderate (GPU-accelerated)	Low (simple weighting function)
Requires prior 3D reconstruction
Resolution improvement	0.2-0.5 Å typical	0.1-0.3 Å typical	0.1-0.2 Å typical

CRYO-EM DATA PROCESSING

Related Terms

Bayesian Polishing is a critical step in the single-particle analysis pipeline. The following concepts are essential for understanding the full context of beam-induced motion correction and high-resolution structure determination.

MotionCor2

A GPU-accelerated software for whole-frame alignment of dose-fractionated movies. It performs global and local motion correction using a patch-based model, generating the aligned averages that serve as input for Bayesian Polishing. Unlike Bayesian Polishing, MotionCor2 operates on frame averages rather than per-particle trajectories.

EXPLORE

Dose Weighting

A computational compensation applied during frame alignment that optimally down-weights later frames in a movie. This accounts for the cumulative loss of high-resolution information caused by radiation damage. Bayesian Polishing extends this concept by modeling the per-particle damage progression within a Bayesian framework rather than applying a global weighting scheme.

Single-Particle Analysis (SPA)

The overarching cryo-EM technique where 2D projections of identical macromolecules in random orientations are computationally combined to reconstruct a 3D density map. Bayesian Polishing is a downstream refinement step within the SPA pipeline, applied after initial 3D reconstruction to correct for residual beam-induced motion that limits resolution.

Gold-Standard FSC

A resolution estimation method that splits particles into two independent half-sets for independent 3D reconstruction. The Fourier Shell Correlation between these half-maps provides an unbiased resolution estimate. Bayesian Polishing typically improves the FSC curve by recovering high-frequency signal previously obscured by uncorrected motion, pushing the reported resolution higher.

RELION

The software package where Bayesian Polishing is natively implemented. RELION (REgularized LIkelihood OptimizatioN) uses an empirical Bayesian approach for cryo-EM structure determination. The Bayesian Polishing algorithm is tightly integrated into RELION's refinement workflow, operating on the output of 3D auto-refinement to perform per-particle motion correction.

Direct Electron Detector (DED)

The hardware enabler for Bayesian Polishing. Cameras like the Gatan K3 or Falcon 4 directly detect electrons with high quantum efficiency and fast readout, enabling dose fractionation into dozens of sub-frames. This frame-level data provides the temporal resolution necessary for Bayesian Polishing to track and reverse per-particle movement trajectories.

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.

Limited slotsGet a Free AI Consultation

How We Work

Custom AI workflows for your Business

One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.