Method for expressing quantitative character and CNV association based on gene interaction network clustering and group sparse learning

Abstract

The invention relates to the field of data mining in bioinformatics, in particular to a synthetic lethal interaction prediction method based on a heterogeneous graph convolutional neural network. The method mainly comprises the steps of (1) collecting breast cancer FPKMRNA-seq information, breast cancer CNV data and breast cancer high-confidence risk gene information; (2) preprocessing breast cancer copy number variation (CNV) and mRNA data by a rank-based method; (3) establishing a gene-gene interaction network based on protein interaction knowledge and a signal channel, and generating a high-density sub-network by using a network clustering algorithm; and (4) constructing a group sparse learning-based model to describe the incidence relation between the subnet and the mRNA expression of the target gene, measuring the ability of the CNVs to predict the expression change of the target gene by using a root mean square error (RMSE), carrying out cross validation by adopting a k-fold cross validation algorithm, and carrying out correlation analysis on the gene expression and the CNV by using a Sperman correlation research method in combination with pathway enrichment analysis.

Description

technical field [0001] The invention relates to the field of data mining in bioinformatics, in particular to a method for associating expression quantitative traits with CNVs based on gene interaction network clustering and group sparse learning. Background technique [0002] Next-generation sequencing (NGS) technologies allow researchers to map the mutational landscape in a specific tumor or even in a pan-cancer fashion. During the development of tumors, different types of mutations have accumulated in the genome, including single nucleotide polymorphisms (SNPs), chromosome fragments, copy number variations (CNVs), gene fusions, etc. [0003] Since different tumor subtypes may adopt different immune evasion pathways, different subtypes have different drug resistance, pathology, and drug responses. The expression of disease-causing and immune genes is regulated at different stages or levels, such as expression quantitative trait loci (eQTLs), CNVs, epigenetic modifications,...

AI Technical Summary

Problems solved by technology

[0004] There are two main shortcomings in the existing calculation methods for studying the regulation mechanism of gene expression: first, the statistical models of the existing methods are relatively simple, and most of them are univariate correlation analysis methods, which do not consider the interaction between multiple factors

Second, the existing research methods are not deep and comprehensive enough to study the role of CNVs

Images