We've developed unique tookits to excel in cryo-EM structure determination.

Graphene supported Grid

GraFuture TM

To overcome the bottleneck in cryo-EM sample preparation, we developed a series of graphene coated grids-GraFuture TM , using them to achieve a better reproducibility and efficiency of Cryo-EM sample preparation.

A big leap can be taken in structure determination once good cryo-EM grids are prepared.

The behavior of molecular embedded in amorphous ice plays an important role in the best achievable resolu- tion for cryo-EM structure determination, leaving specimen preparation to be a major bottleneck limiting in this method.

Aiming to improve reproducibility and controllability, we developed a facile and robust strategy to use reduced graphene oxide(RGO) membrane as supporting film in cryo-EM specimen preparation. This grid has been used in our platform and showed great power to reconstruct biomolecules (including sub-100-kDa molecules), even at near-atomic resolution that can not be reached by conventional used grid.

GraFuture TM GO

Graphene Oxide Support Film TEM Gird

GraFuture TM RGO

Reduced Graphene Oxide Support Film TEM Gird

GraFuture TM SCG

Single-Crystalline Graphene Support Film TEM Grid

GraFutureTM : offering promise for challenging samples that are inaccessible by cryo-EM before.

Compared with conventional supporting films, EM grids coated with our RGO membrane was characterized with several superior properties.

Decreased interlayer space results in thinner ice which is just thick enough to support the particles. Lower back- ground noise significantly improves the image quality.

Improved biocompatibility and nice particle-absorption ability not only enhances particle concentration, but also alleviates air water interface problem which is hypothesized to cause denaturation and preferred orientation.

Enhanced electrical/thermal conductivity contributes to no visible beam-induced footprints and minimized radiation damage.

High mechanical strength means adequate strength to support the ice layer and reduced beam-induced motion.

Using our grids, boundaries are continuously being pushed in structure determination.

Successfully solved the structure of sub-100 kDa macromolecules at 2.6 Å resolution.

Applicable to cryo-sample preparation for proteins with low concentration(<100 nM).

Absorb particles onto the surface of graphene supporting film, reducing air-water interface denaturation.

Better support particles with a more balanced orientation distribution.

Structure Modeling with Deep Neural Network


We are developing a new differentiable neural network based method (modelSMART) to directly identify the 3D atomic model from cryo-EM density map in a few hours with- out any human interference.

Model building remains a bottleneck as the last step in structure determination, although the progress is becoming more and more automated.

Manual operation is time-consuming and inevitably error-prone. Significant incorporation of automation has re- sulted in the ability to obtain near-atomic resolution reconstructions within hours to days. However, molecular model building is difficult to automate since it relies substantially on human expertise, which limits scalability for use by wider community. A model that represents the atomic coordinates of each amino acid is constructed and fitted into the 3D electron density volume as the final step. Specific amino acids to a density volume are manually assigned by structural biology experts with the help of 3D visualization tools. Attempts to automate this process are far from satisfactory given to the accuracy, coverage, time-consuming and requirement of human intervention.

AI-driven: faster and more accurate than existing automated solutions without any human intervention.

As part of the push to improve efficacy and performance of model building, We proposed a learning-based method from a totally novel perspective. The power of deep neural network is leveraged to develop a novel framework for end-to-end atomic model building from Cryo-EM density map.

These endeavor for better model building is expected to fuel the rapid growth of cryo-EM atomistic structures. We hope our exploration will facilitate the integration of machine learning and more collaboration with experts from fields outside of traditional structural biology.

  • This is the first attempt to formulate molecular structure determination from Cryo-EM density volumes with a deep learning approach.
  • We adapt a novel 3D network architecture for atomic model building in density volumes, and proposed a Cryoformer block for better performance.
  • We carefully designed a way to bridge Cryoformer and AlphaFold to achieve Cryo-EM density-based protein complex structure prediction.


As a novel crystallographic method, MicroED exploits the advantage that electrons interact much more strongly with material and posit considerably less damaging energy into a crystal. Electron diffraction data can be collected from extremely small nanocrystals at a low dose, which not only benefits structure research of organic compound, but also for crystallizing problematic proteins. With this new modality in cryo-EM, novel structures considered untraceable before were successfully determined today.

However, the special requirement for radiation-tolerated movie-mode camera and the lack of automated data collection method increases the barrier to the utilization of MicroED. In SHUIMU, we used our stage-camera synchronization scheme and automatic data collection software-eTasED for conventional Cryo-EM, making it possible to solve small molecule and protein structures in ultrahigh-resolution using a low-end electron microscope and improving the general applicability of MicroED.

Schematic diagram of the stage-camera synchronization

Benefiting from eTasED, microED can be realized on a conventional cryo-EM system without any modification.

  • Tailored for single-frame camera. The stage tilting and sample illumination are activated and inactivated at predefined time points so that they are only activated within the effective exposure step of an exposure cycle.
  • Available for movie-mode camera. With a movie-mode camera, the exposure time can be set much longer than that of a single-frame camera, such that each tilting-exposure cycle covers a large or even entire range of the tilting angle.
  • Compatible with 120kV TEM. Ultrahigh-resolution diffraction data of peptide nanocrystals is collected on 120-kV electron microscopes, resulting in resolution up to ∼0.60 Å with unambiguous assignment of nearly all hydrogen atoms.

With the application of e-TasED, structures not only for small molecules but also for peptides and proteins can be realized by microED in SHUIMU.

Proteinase K
1.50 Å, 29.05 kDa

0.65 Å, 0.66 kDa

0.65 Å
(small molecule)

Together We Can Make
Game-Changing Discoveries.

Partner with us