Gene module mining method, device, and computer equipment based on graph neural network

Abstract

The invention discloses a gene module mining method, device and computer equipment based on a graph neural network. Wherein, the method includes: constructing a gene co-expression network according to the gene expression profile data, and based on the constructed gene co-expression network, configuring a community membership matrix through a graph neural network, and a community membership matrix based on the configuration , generate known modules by setting thresholds. Through the above method, it is possible to configure the community affiliation matrix through graph neural network representation learning, and then generate known modules by setting thresholds, so that multiple genes with dense connections can be assigned to the gene module mining results. different modules.

Description

technical field [0001] The present invention relates to the technical field of gene module mining, in particular to a gene module mining method, device and computer equipment based on a graph neural network. Background technique [0002] In related technologies, with the rapid development of high-throughput biological experiment technology, especially the development of gene chip and next-generation sequencing technology, the biological data of the whole genome is growing explosively. Through the analysis of these experimental data, some different types of networks can be obtained, such as gene expression regulation network, protein interaction network, and transcription regulation network. The molecular biological network reflects the interaction relationship of biomolecules at the system level, so it helps researchers to understand how various biomolecules interact in biological cells to a certain extent, and then perform the complete processing process of biological funct...

AI Technical Summary

Problems solved by technology

[0004] In the mining task of gene expression profile data, clustering only focuses on the co-expression of genes in all samples, while ignoring the local expression relationship that may only exist in some samples, and clustering can only divide genes into a single module However, more and more evidences have shown that a single gene is often associated with multiple functions or metabolic pathways. How to mine such overlapping modules is also a difficult point in gene module mining, which makes it impossible to realize the results of gene module mining. Allows multiple genes with dense connections to be assigned to different modules

Images