This release includes cryoDRGN-AI Ab Initio reconstruction, launches MotionRefine CC (alpha), expands the embedded volume viewer across multiple job types, and delivers a redesigned project overview header, alongside performance improvements and workflow refinements.
🚀 New Features
cryoDRGN Ab Initio now available
CryoCloud already supported cryoDRGN for neural-network–based reconstruction of heterogeneous cryo-EM structures. With this release, we now also support cryoDRGN-AI Ab Initio, enabling neural ab initio reconstruction of challenging cryo-EM and cryo-ET datasets without requiring a starting model.
Reconstructs continuous distributions of 3D density maps and models per-particle heterogeneity using deep neural networks.
cryoDRGN-AI (Levy et al., Nature Methods 2025) Performs ab initio reconstruction using neural representations combined with exhaustive search and gradient optimisation, which can be particularly powerful for heterogeneous and difficult datasets.
In our internal benchmarking (EMPIAR-11918, 39 kDa complex), cryoDRGN-AI Ab Initio significantly improved over a RELION initial model approach (see above).
We invite you to try cryoDRGN-AI Ab Initio directly in CryoCloud and explore how it performs on your datasets!
(ALPHA) MotionRefine CC
We have released our own GPU-accelerated MotionRefine CC algorithm, currently in alpha. MotionRefine CC is intended for polishing workflows (for training, please continue using the RELION legacy implementation).
In preliminary benchmarks, MotionRefine CC shows 2-3× faster performance, resulting in reduced compute time per job and lower processing costs.
As this feature is in alpha, we are continuing to validate performance across diverse datasets and workflows. While stable for use, we welcome feedback as we further optimise and expand its capabilities.
Embedded Volume Viewer Expansion
You asked — we answered! We are progressively adding embedded output visualisation directly to results pages. In this release, embedded viewers were added to:
If there are additional result pages you would like visualisation added to, let us know at support@cryocloud.io.
Project View Overhaul
The project page header has been redesigned to provide clearer insights into data processing at a glance.
On the bottom-left side, you can now see: 1) total number of jobs, 2) total storage usage, and 3) total compute hours — broken down per job type.
The right-hand summary panel highlights the best results across the five stages of a single-particle analysis project: Pre-processing, Class2D, Class3D, Refine3D, and Post-processing. Clicking a highlighted result takes you directly to the corresponding job.
✨ Improvements
Performance
Up to 4× faster CTF Estimation jobs, driven by optimised denoising and downscaling
Faster and more reliable STAR file reading, including support for multi-GB STAR files
AutoClass2D
AutoClass2D can now accept particles.star from Class2D or Select jobs, enabling pre-filtered particle input (e.g. RELION Class Ranker score filtering).
X-ray Crystallography Pipeline
Space group and unit cell parameters are now exposed in the Pipedream job.
Workflows
Workflows now pause correctly at InteractiveSelect jobs.
User Interface
Unit symbol (Å) added to particle diameter label.
🐛 Bug Fixes
AutoClass2D now gracefully handles empty classes in the parent Class2D job
Dataset search filtering and multi-dataset selection were restored when starting projects from workflows
Fixed the labels column filter to include Targets
Various stability and validation fixes across jobs and workflows
Missed an update?
View all CryoCloud release notes and version history in one place.
You’re receiving this email because you have an active CryoCloud account. If you don’t wish to receive future release updates, you can unsubscribe at the bottom of this email.
