Scientists propose a new method for cryo-electron tomography 3D imaging target recognition

On May 22, Zhu Ping’s research group at the Institute of Biophysics, Chinese Academy of Sciences published a paper in the international academic journal Nature Communications. In this paper, the researchers propose a method for direct observation and identification of high signal-to-noise ratio of in-situ structural features and dynamic conformations of target molecules in cryo-electron tomography 3D imaging, and named REST (REstoring the Signal in Tomograms).

Cryo-electron tomography can obtain in situ three-dimensional structures of biological macromolecules with nanoscale resolution in cell and tissue samples, but due to the extremely low signal-to-noise ratio and irreversible information loss in cryo-electron tomography, it is difficult for researchers to obtain the real information of target particles (ground truth) required in the deep learning process, which makes the use of neural networks and deep learning techniques to identify target macromolecular proteins in electron tomography with great challenges.

In order to solve the above technical bottlenecks, the newly published research paper of Zhu Ping’s research group proposes and implements two training strategies. In strategy 1, the researchers select a small number of particles from the original data for subunit averaging, use the average result as the “ground truth” of training, and establish a training pair with the original particles. In the second strategy, the researchers artificially add different degrees of noise and dynamic conformational changes to the high-quality “ground truth” density map to simulate the low signal-to-noise ratio and macromolecular structure heterogeneity in the real data, and establish a mapping and training set for the high-noise, dynamically changing low-mass particle density map obtained by the simulation with the high-quality density map.

After establishing the above training set and deep learning strategy, the researchers used the deep learning network to learn and train the training set, and transferred the trained model and acquired knowledge to the original data to recover the information of the target protein particles.

REST methodology, flow, and training strategy

It is found that using the above strategies, REST methods will have a wide range of application values and prospects in various tasks related to cryo-electron tomography, such as restoring the clear signal of the target protein (such as identifying and extracting particles in a noisy background), segmenting target features, identifying the dynamic or flexible structure of the target protein, obtaining a density without missing information as an initial model, and assisting electron tomography with sub-unit average (STA).

Zhang Haonan, a doctoral student in Zhu Ping’s research group at the Institute of Biophysics, Chinese Academy of Sciences, and Li Yan, an associate researcher, are the co-first authors of the paper, and researcher Zhu Ping is the corresponding author of the paper. The research work was supported by the National Natural Science Foundation of China, the key research and development projects of the Ministry of Science and Technology, and the Strategic Leading Science and Technology Project of the Chinese Academy of Sciences (Class B). (Source: Meng Lingxiao, China Science News)

Related paper information:

Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button