Structural Biology

Cryo-EM Single-Particle Analysis: From Raw
Micrographs to 3D Reconstruction

BioMate runs the full CryoSPARC single-particle analysis pipeline on NVIDIA A100 GPU instances — motion correction, CTF estimation, particle picking, 2D/3D classification, and gold-standard FSC refinement — and returns an EMDB-ready density map with resolution report.

Try free See structural biology services
Pipeline stages

End-to-end SPA pipeline — every stage automated

BioMate orchestrates each stage of the CryoSPARC single-particle analysis pipeline, streaming real-time progress back to the UI. No manual job chaining, no intermediate file management.

Motion Correction

Beam-induced motion is corrected across all movie frames to produce sharp, averaged micrographs. BioMate uses MotionCor2 with dose-weighting for optimal contrast at high spatial frequencies.

MotionCor2dose-weightinggain reference applied

CTF Estimation

Contrast transfer function parameters (defocus, astigmatism, phase shift) are estimated per micrograph. Micrographs outside the acceptable CTF quality range are automatically excluded from further processing.

CTFFIND4CryoSPARC CTFquality filtering

Particle Picking

Particles are picked automatically using CryoSPARC blob picker for initial runs, with TOPAZ neural network picking available for low-contrast or difficult particles (membrane proteins, small complexes <100 kDa).

CryoSPARC blob pickerTOPAZ neural pickerinspect picks QC

2D Classification

Iterative 2D classification removes junk particles, ice contamination, and overlapping particles. Class averages are displayed for review, and bad classes are automatically excluded before 3D reconstruction.

CryoSPARC 2Diterative refinementautomatic junk removal

3D Ab Initio Reconstruction

Initial 3D model built without a reference structure, avoiding model bias. Multiple classes generated simultaneously to identify conformational heterogeneity or preferred orientations.

CryoSPARC ab initiono template requiredmulti-class

3D Refinement

Homogeneous or heterogeneous refinement to high resolution. Non-uniform refinement available for flexible assemblies. Local refinement for focused reconstruction of specific domains.

homogeneous refinementnon-uniform refinementlocal refinement

Post-processing & Sharpening

B-factor sharpening, local resolution estimation, and gold-standard FSC 0.143 criterion applied to report final resolution. Map validated and packaged for EMDB deposition.

B-factor sharpeningFSC 0.143local resolutionEMDB-ready
Infrastructure

GPU-accelerated on AWS Batch

CryoSPARC requires GPU compute for every stage beyond motion correction. BioMate provisions NVIDIA A100 instances on demand through AWS Batch — no GPU cluster to manage, no license headaches.

NVIDIA A100 · 40 GB VRAM per GPU · NVMe local storage

Each cryo-EM job runs on dedicated A100 instances with fast NVMe scratch storage for micrograph stacks. Instances spin up when the job is submitted and terminate when it completes — you pay only for the compute you use, not idle GPU time.

4–12 h
typical job time (1k–5k micrographs)
A100
NVIDIA GPU — 40 GB VRAM
Results

What you receive

Every cryo-EM run produces a complete output package. Files are previewed directly in BioMate and downloadable in their native format.

.mrc

Refined density map

Sharpened, masked map at gold-standard FSC 0.143 resolution. EMDB-submission format.

.mrc

Half-maps

Unmasked half-maps for independent validation and downstream local resolution estimation.

PNG

FSC & B-factor curves

Gold-standard FSC plot with 0.143 cutoff annotated. B-factor and local resolution heatmaps.

PNG

2D class averages

Final accepted 2D class averages showing particle views across all orientations.

CSV

Particle metadata

Star file / CS file with all particle coordinates, CTF parameters, and refined pose angles.

DOCX

Methods report

Publication-ready methods section with all parameters, software versions, and resolution estimate with confidence interval.

FAQ

Common questions about cryo-EM in BioMate

What input files does cryo-EM analysis require?

Raw micrograph movies in .mrc or .tiff format from your cryo-EM microscope. BioMate auto-detects pixel size, voltage, and spherical aberration coefficient if embedded in the file header. Gain references and defect maps can be uploaded optionally for motion correction.

Which software does BioMate use for cryo-EM?

CryoSPARC for 2D/3D classification, ab initio model building, and homogeneous and heterogeneous refinement. MotionCor2 for beam-induced motion correction, and CTFFIND4 for contrast transfer function estimation. TOPAZ neural network particle picker is available for low-contrast particles.

How long does a cryo-EM run take?

Typically 4–12 hours for a standard single-particle dataset of 1,000–5,000 micrographs on NVIDIA A100 GPU instances on AWS Batch. Motion correction and CTF estimation are parallelized across micrographs. CryoSPARC refinement scales with particle count and box size.

What resolution can BioMate achieve?

Resolution depends entirely on data quality — particle size, sample preparation, ice thickness, and microscope calibration. BioMate applies the gold-standard FSC 0.143 criterion and reports the estimated resolution with its confidence interval. Near-atomic resolution (<3 Å) is achievable for well-behaved samples; BioMate does not over-report resolution.

Get started

Process your cryo-EM data on GPU today

Upload your micrograph movies and describe your sample. BioMate runs the full CryoSPARC SPA pipeline and returns a density map with resolution report — no local GPU cluster required.

Try free →