Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Implementation method of single particle image clustering for cryo-EM based on graph convolutional autoencoder

A convolutional self-encoding and cryo-electron microscope technology, which is applied in the field of cryo-electron microscope single-particle image clustering, can solve problems such as large noise influence, and achieve the effect of improving robustness and clustering results

Active Publication Date: 2022-07-15
SHANGHAI JIAO TONG UNIV
View PDF8 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] Aiming at the defect that the existing technology is greatly affected by noise, the present invention proposes a cryo-electron microscope single-particle image clustering method based on graph convolutional self-encoder, using a networked similarity measurement method and a local linear embedding extraction method, Combined with the hidden layer node features of the autoencoder, the network structure information and the image feature information of the node itself are simultaneously learned, which improves the robustness of the clustering and significantly improves the image quality

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Implementation method of single particle image clustering for cryo-EM based on graph convolutional autoencoder
  • Implementation method of single particle image clustering for cryo-EM based on graph convolutional autoencoder
  • Implementation method of single particle image clustering for cryo-EM based on graph convolutional autoencoder

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0022] In this embodiment, the activation functions used in the encoding layer and the decoding layer of the graph convolutional autoencoder are both sigmoid functions, the optimizer adopts the Adam optimizer, and the inner product decoder adopts the cross-entropy of the reconstruction network and the input network as the loss function , while the inverse graph convolutional decoder uses the squared error of the reconstructed node features and the input node features as the loss function. Both the graph convolutional encoder and the inverse graph convolutional encoder are two-layer graph convolutional neural networks, in which the number of neurons in the two layers is set to 32 and 16, respectively.

[0023] like figure 2 Shown is the real cryo-electron microscope single particle image of GroEL protein in the GroEL real image set, which is 4096 128×128 cryo-electron microscope single particle images of GroEL protein with a structure with D7 symmetry. The goal of this embodi...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

A method for realizing single-particle image clustering in cryo-EM based on graph convolution autoencoder, which generates KNN network by calculating the image similarity matrix of cryo-EM single-particle image set; The particle image is dimensionally reduced, the KNN network and the image feature matrix are input into the encoder in the graph convolution self-encoder, and the high-dimensional node features are embedded into the low-dimensional hidden layer space to obtain a low-dimensional hidden layer structure. After the point feature, the K-means clustering process is performed by the decoder in the graph convolution self-encoder to obtain the clustering result of the cryo-EM single particle image, and finally the images of each cluster are averaged to obtain the final class. Average image. The invention uses the networked similarity measurement method and the local linear embedding extraction method, and combines the hidden layer node features of the self-encoder to learn the structure information of the network and the image feature information of the node itself at the same time, so that the robustness of the clustering is improved. Boosted to significantly improve image quality.

Description

technical field [0001] The invention relates to a technology in the field of image processing, in particular to a method for realizing clustering of cryo-electron microscope single-particle images based on a graph convolution self-encoder. Background technique [0002] Cryo-electron microscopy is a technique in which macromolecules to be detected are placed in an ultra-low temperature environment and then observed with an electron microscope. The three-dimensional model of the biomacromolecule can be obtained by reconstructing the two-dimensional image obtained by the electron microscope. Cryo-EM single particle image reconstruction is the most commonly used cryo-EM reconstruction method. In the detection process of cryo-electron microscopy, in order to ensure that the activity of biological macromolecules is not affected by electron radiation as much as possible, the electrons used by the electron microscope must be kept at a very small dose, so the noise of single-particl...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06V10/762G06V10/44G06K9/62G06V10/82G06N3/04G06N3/08
CPCG06N3/08G06V10/44G06N3/048G06N3/045G06F18/213G06F18/23213G06F18/22
Inventor 蔡嘉鸣沈红斌
Owner SHANGHAI JIAO TONG UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products