CryoNet.Refine: Fully automated cryo-EM model refinement through physics-informed deep learning

CryoNet.Refine improves an input structural model directly against a cryo-EM density map in one streamlined step, making it practical to polish candidate models quickly before downstream analysis.

Upload a cryo-EM density map and one initial structural model.

Use the reported map resolution as an explicit refinement condition.

Receive a refined structure suitable for downstream inspection and modeling.

Start Refinement

Demo Video

Full frame playback enabled

This video demonstrates the online CryoNet.Refine experience. It is embedded here so users can immediately understand what the service does before they submit a job.

Submit A Refinement Job

Provide a cryo-EM density map, an initial atomic model, and the target map resolution. The server will generate a refined structure constrained by the density.

Cryo-EM density map *

Choose file select *.mrc file (<= 2GB)

Atomic Model *

Choose file select *.cif/*.pdb file

Resolution (Å)*

Target Name

Email*

Example 1: Cryo-EM structure of Cas13(crRNA) i

Cryo-EM map file: 21367.mrc

Atomic model file: 21367_af3.cif

Resolution (Å): 3.2

Example 2: Concentrative nucleoside transporter CNT3 i

Cryo-EM map file: 0775.mrc

Atomic model file: 0775_af3.cif

Resolution (Å): 3.6

Example 3: Medicago truncatula HISN5 protein i

Cryo-EM map file: 12938.mrc

Atomic model file: 12938_af3.pdb

Resolution (Å): 2.4

Reference

CryoNet.Refine: A One-step Diffusion Model for Rapid Refinement of Structural Models with Cryo-EM Density Map Restraints , Fuyao Huang*, Xiaozhu Yu*, Kui Xu#, and Qiangfeng Cliff Zhang#, (2026), International Conference on Learning Representations, ICLR.

Code

Github