For structural-biology biotechs & cryo-ET cores

Cryo-ET pipelines on your AWS — no academic-cluster waitlist.

RELION-5, AreTomo, Warp/M, IsoNet, MemBrain, MotionCor3, CTFFIND4, Topaz — verified containers, multi-GPU, Slurm-orchestrated. Multi-TB tilt-series stay in your VPC. We containerize the new methods that drop quarterly so your team doesn't have to.

RELION-5 AreTomo Warp / M IsoNet denoising MemBrain particle picking MotionCor3 + CTFFIND4

Plans from $99/mo BYOC · free 30-day pilot · How we compare to CryoCloud →

Validated result

3.99 Å subtomogram average. $2.30 in GPU spot.

We ran the full RELION-tomo STA pipeline on EMPIAR-10164 (immature HIV-1 Gag) on Clusterra: AreTomo2 tilt-series alignment → RELION-tomo particle extraction → gold-standard 3D refinement. Gold-standard FSC resolution: 3.99 Å. Total GPU cost on 4× T4 spot: ~$2.30.

Read the full cryo-ET case study →

Cryo-ET toolchain

Verified containers for the in-cell structural-biology stack.

Pinned versions, GPU-tuned, multi-node where applicable. The exact things you'd install on a fresh academic cluster, minus the install.

Why BYOC for cryo-ET

Multi-TB tilt series are not someone else's problem.

01 / Data gravity

5–10+ TB per microscope session. Don't upload it to a vendor cloud.

Tilt-series and reconstructed tomograms stay in your S3 bucket. Cross-account IAM, your KMS keys, your audit trail. We never see the data.

02 / No academic queue

RELION 3D refinement at p4/p5 GPU scale, on demand.

Skip the 2-week academic-cluster waitlist. Pay only for the GPU-hours you use. Multi-GPU NCCL + EFA for large refinements; spot checkpoint-restart for long jobs.

03 / Frontier tools, fast

The new method that dropped last quarter is already containerized.

IsoNet, MemBrain, AreTomo, Warp/M — we track the field. When the next denoiser or picker drops, we container-build and ship it. You don't maintain a custom RELION install with five patches.

Who this is for

A real fit, honestly.

Fit

You probably want to talk to us if …

· 5–50 person structural-biology biotech (cellular cryo-ET, drug-target structure)
· Or a cryo-EM core facility serving industry users
· Running RELION / AreTomo / Warp pipelines on aging on-prem GPUs or shared academic clusters
· Need GPU + fast scratch for tomogram reconstruction and subtomogram averaging
· Want a managed catalog of frontier OSS tools (IsoNet, MemBrain, etc.) without maintaining containers yourself

Not fit

Probably not for you if …

· Your pipeline is fully on cryoSPARC and the license is happy
· You have an in-house cryo-EM IT team running your own cluster
· You only do SPA (single-particle), not cryo-ET — the wedge here is in-cell tomography
· You need SOC 2 / FedRAMP in month one
· Microscope time is your only bottleneck and processing turnaround doesn't matter

Published benchmarks

Real numbers on real data.

All runs on Clusterra — managed Slurm, BYOC, spot EC2.

relion-cryoet-hiv-sta

Cryo-ET STA — 3.99 Å on EMPIAR-10164

Immature HIV-1 Gag subtomogram average. AreTomo2 tilt-series alignment → RELION-tomo gold-standard refinement. 4× T4 spot, ~$2.30 total. Full end-to-end STA pipeline.

Read →

relion-cryoem-3dclass

Cryo-EM SPA — RELION-5 3D classification

Multi-GPU RELION-5 3D classification and refinement on a public EMPIAR dataset. Same managed Slurm stack; demonstrates the SPA pipeline for teams running both SPA and cryo-ET.

Read →

sarek-parabricks

Multi-TB I/O under spot — WGS at scale

30× WGS in 2h 32m for ~$2.59/sample. Parabricks GPU alignment + DeepVariant. F1=0.9966 against GIAB. Demonstrates large-input pipelines under spot preemption.

Read →

$ sbatch cryo-et-pipeline --tilt-series sample.mrc # RELION-5 + AreTomo2 + IsoNet + MemBrain on your AWS.

Run your tilt series on the same stack.

Free 30-day pilot. Your data in your account, RELION-5 + AreTomo2 running before the scoping call ends — no charge. Plans from $99/mo BYOC.