Resources · Data-Processing Pipeline

Cryo-EM Processing Pipeline

The end-to-end single-particle workflow our trainees follow, from raw movie frames to a refined, interpretable 3D reconstruction. Adapted from the SBGrid NIH R25 Train-the-Trainer curriculum, co-led by our lab.

Single-particle cryo-EM

Cloud-based (SBCloud)

Trainee-led, reproducible

Workflow steps

A complete image-processing pipeline carrying raw detector movies through to a filtered, interpretable density map — the same sequence taught in SBGrid's two-day Single-Particle CryoEM Data Processing curriculum.

Single-particle cryo-EM data processing flow: project and data setup, preprocessing, particle picking, 2D classification, initial model, 3D classification, 3D refinement, visualization and model building

Workflow overview & data organization

Establish a reproducible project structure and review the full pipeline before processing begins, so raw data, intermediate files, and final outputs stay traceable.

cryoSPARC RELION

Data import & organization

Import raw movie stacks from the detector, verify metadata (pixel size, dose, defocus range), and organize sessions for downstream processing.

cryoSPARC RELION

Motion correction & CTF estimation

Correct beam-induced motion across movie frames and estimate the contrast transfer function per micrograph to recover high-resolution signal.

MotionCor2 CTFFIND4 cryoSPARC

Particle picking & extraction

Identify individual particle projections on each micrograph (template- or neural-net-based picking) and extract boxed particle images.

Topaz cryoSPARC

2D classification & particle selection

Sort particles into 2D class averages to remove junk picks, ice contamination, and damaged particles before 3D processing.

cryoSPARC RELION

Initial model generation

Generate a low-resolution ab initio 3D model directly from the cleaned 2D particle stack, with no external structural prior required.

cryoSPARC

3D classification & refinement

Separate conformational or compositional heterogeneity into distinct 3D classes, then refine the best class to a high-resolution consensus reconstruction.

RELION cryoSPARC

Local resolution estimation & filtering

Map local resolution across the reconstruction and apply resolution-aware filtering/sharpening so weaker, flexible regions aren't over-sharpened.

ResMap DeepEMhancer

Visualization & interpretation

Inspect the final density map in ChimeraX to assess map quality and hand off to model building (see the companion Model Building & Validation Workflow).

ChimeraX

Worked training example

EMPIAR‑11422 — full-length dimeric ClbP

The SBGrid Train-the-Trainer curriculum runs this exact pipeline on a real Titan Krios dataset, EMPIAR‑11422, which produced the cryo-EM structure of full-length dimeric ClbP (PDB 7UL6; EMD‑26593), originally published by Velilla and colleagues in 2022. Trainees work with authentic experimental data end-to-end rather than a synthetic tutorial dataset.

References

Primary sources for this workflow. Each link goes to the original resource so you can confirm details directly.

SBGrid NIH R25 Training — CryoEM & CryoET Data Processing WorkshopsProgram overview, curriculum topics, and Train-the-Trainer model (SBGrid Consortium, Harvard Medical School).

EMPIAR‑11422Public raw-movie dataset used in the workshop's single-particle curriculum.

PDB 7UL6Deposited atomic model resulting from this dataset.

EMD‑26593Deposited cryo-EM density map for full-length dimeric ClbP.

Velilla et al., 2022, Nature Chemical BiologyOriginal publication describing the ClbP structure and its biological significance.

SBCloudCloud-based training environment used to run the pipeline (curated software, CPU/GPU, storage, visualization).