CN-116434827-B - Particle screening method, device, equipment and storage medium in freeze electron microscope imaging

CN116434827BCN 116434827 BCN116434827 BCN 116434827BCN-116434827-B

Abstract

The application relates to a particle screening method, a device, equipment and a storage medium in a freeze electron microscope imaging, wherein the method comprises the steps of dividing a protein reconstruction final particle data set into half data sets based on a division strategy in the three-dimensional reconstruction of the freeze electron microscope, respectively carrying out three-dimensional reconstruction on all particles in each half data set to obtain a first protein density map and a second protein density map so as to screen all particles in the corresponding half data sets to obtain particles meeting preset conditions, carrying out iteration on the three-dimensional reconstruction and the particle screening until the preset iteration times are reached, obtaining a first protein density map and a second protein density map with different rounds and the resolution ratio of the first protein density map and the second protein density map, and further determining the optimal screening proportion of the final particle data set. Therefore, the problems that the existing particle screening technology is too dependent on a screening criterion formulated by human, the screening time is long, the accuracy of a screening algorithm is low, a large number of particles with low signal to noise ratio are difficult to screen rapidly and effectively are solved.

Inventors

BAO CHENGLONG
ZHU JIANYING
ZHANG QI
HU MINGXU
Shi Zuoqiang

Assignees

清华大学

Dates

Publication Date: 20260512
Application Date: 20230322

Claims (6)

1. The particle screening method in the freeze electron microscope imaging is characterized by comprising the following steps of: Dividing a protein reconstruction final particle dataset under the imaging of a frozen electron microscope into two half-data sets with the same particle number based on a preset dividing strategy; Respectively carrying out three-dimensional reconstruction on all particles in each half data set to obtain a first protein density map, a second protein density map and corresponding resolution; performing particle screening on all particles in the corresponding semi-data set based on the first protein density map and the second protein density map to obtain particles meeting preset conditions, iterating three-dimensional reconstruction and particle screening until the preset reconstruction-screening iteration times are reached, obtaining the first protein density map, the second protein density map and the corresponding resolution of each iteration, and determining the optimal screening proportion of a final particle data set; The step of screening the particles in the corresponding semi-data set based on the first protein density map and the second protein density map to obtain the particles meeting the preset condition includes: calculating features of the all particles based on the first and second protein density maps and the corresponding semi-data sets; Sorting the features of all particles in a descending order, and removing M particles with the largest value from each pair of half data sets, wherein M is a positive integer; the characteristic formula of each particle is as follows: Wherein, the In order to truncate the operator in the fourier domain, In order to obtain a map of the density of the protein, Is a composite observation operator of particles in the frozen electron microscope imaging, In the form of a discrete fourier transform, In order to cut off the frequency of the signal, The particles in the set of semi-data are for each of the pairs.
2. The method of claim 1, further comprising, prior to dividing the protein reconstruction final particle dataset under the cryo-electron microscopy imaging into two of the same particle count pair of half datasets based on the preset division strategy: The rotation angle, translation amount, and CTF imaging parameters of each particle in the final particle dataset are estimated for the protein reconstruction.
3. A particle screening device in cryoelectron microscopy imaging, comprising: The dividing module is used for dividing the protein reconstruction final particle data set under the imaging of the frozen electron microscope into two half data sets with the same particle number based on a preset dividing strategy; The reconstruction module is used for respectively carrying out three-dimensional reconstruction on all particles in each half data set to obtain a first protein density map, a second protein density map and corresponding resolution; The screening module is used for screening all particles in the corresponding semi-data set based on the first protein density map and the second protein density map to obtain particles meeting preset conditions, iterating the three-dimensional reconstruction and the particle screening until the preset reconstruction-screening iteration times are reached, obtaining the first protein density map, the second protein density map and the corresponding resolution ratio of each iteration, and determining the optimal screening proportion of the final particle data set; wherein, the screening module includes: a computing unit for computing features of all particles based on the first and second protein density maps and the corresponding semi-data sets; the rejecting unit is used for sorting the characteristics of all particles in a descending order and rejecting M particles with the largest value from each half data set, wherein M is a positive integer; the characteristic formula of each particle is as follows: Wherein, the In order to truncate the operator in the fourier domain, In order to obtain a map of the density of the protein, Is a composite observation operator of particles in the frozen electron microscope imaging, In the form of a discrete fourier transform, In order to cut off the frequency of the signal, The particles in the set of semi-data are for each of the pairs.
4. A device according to claim 3, further comprising: And the estimation module is used for estimating the rotation angle, the translation amount and the CTF imaging parameters of each particle in the protein reconstruction final particle data set before dividing the protein reconstruction final particle data set under the frozen electron microscope imaging into the two half data sets with the same particle number based on the preset division strategy.
5. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor executing the program to implement the particle screening method in cryo-electron microscopy imaging as defined in any of claims 1-2.
6. A computer readable storage medium having stored thereon a computer program, characterized in that the program is executed by a processor for implementing a particle screening method in cryo-electron microscopy imaging as claimed in any of claims 1-2.

Description

Particle screening method, device, equipment and storage medium in freeze electron microscope imaging Technical Field The application relates to the technical field of protein particle screening, in particular to a particle screening method, a device, equipment and a storage medium in a freeze electron microscope imaging. Background The basic flow of the method is as follows: 1. Purifying and freezing the target protein sample; 2. shooting a sample frozen on an ice layer by using a high-energy electron beam to obtain a large number of two-dimensional particle projection images; 3. The three-dimensional conformation of the protein is reconstructed from the two-dimensional particles using the maximally expected algorithm in machine learning. However, due to interference of sample preparation, imaging mechanism, mathematical modeling and the like, a large number of particles influencing the accuracy of three-dimensional reconstruction exist in the two-dimensional particle data set, so that designing an effective algorithm for particle screening has become an important problem in the field of frozen electron microscopy. At present, particle screening in the field of frozen electron microscopy is completed through multi-round clustering, including two-dimensional classification and three-dimensional classification of particles, in the actual operation process, after a user performs multi-round two-dimensional and three-dimensional classification on the particles, the user roughly selects a large number of particles in a micrograph (Micrograph), then error particles such as ice crystals and carbon films are removed through multi-round two-dimensional classification, reconstruction and three-dimensional classification are repeated, and all the particles in the category based on visual observation of bad particles are discarded, and the particles are classified into the category with a homogeneous and uniform structure through three-dimensional classification. However, in the prior art, in the two-dimensional and three-dimensional classification process, the determination of the class standards of good and bad particles is too dependent on the subjective standards of users, and the standards among different users are not necessarily consistent, so that in order to ensure that the particles with higher quality are not screened, the screening of the particles is quite conservative, the particles with the class of protein shape are prone to not removing blur, in addition, repeated classification and reconstruction require a great deal of time, especially when three-dimensional classification is used for selecting homogeneous particles, the user is often required to repeatedly perform operations such as classification, reconstruction and particle combination for a plurality of months, meanwhile, the signal-to-noise ratio of the particles in a freezing electron microscope is extremely low, the clustering algorithm is easy to misclassify the particles, if the good particles are screened as bad particles, the quantity of the particles for reconstruction is reduced, the final reconstruction resolution is reduced, if the bad particles are kept as good particles, the bad particles are mixed with a certain proportion of the particles for reconstruction, and the quality of the reconstructed protein structure is reduced. In summary, the existing particle screening technology is too dependent on the manually formulated screening criteria, and has the disadvantages of long screening time, low accuracy of the screening algorithm, and difficulty in rapidly and effectively screening out a large number of particles with low signal to noise ratio. Disclosure of Invention The application provides a particle screening method, a device, equipment and a storage medium in a frozen electron microscope imaging, which are used for solving the problems that the existing particle screening technology is too dependent on a manually formulated screening criterion, the screening time is longer, the accuracy of the screening algorithm is lower, a large number of particles with low signal to noise ratio are difficult to screen rapidly and effectively, and the like. The embodiment of the first aspect of the application provides a particle screening method in a freeze electron microscope imaging, which comprises the following steps of dividing a protein reconstruction final particle dataset under the freeze electron microscope imaging into two half-data sets with the same particle number based on a preset division strategy, respectively carrying out three-dimensional reconstruction on all particles in each half-data set to obtain a first protein density map, a second protein density map and corresponding resolution, carrying out particle screening on all particles in the corresponding half-data set based on the first protein density map and the second protein density map to obtain particles meeting preset conditions, and carrying out iteration on the