Correlation Analysis And Prediction Based On Phenotype Gene Double Coupling Network

After the sequencing of the human genome project,life scientists have stepped into the research field of gene function.It has become an important topic to solve the relationship between gene and disease.The vast majority of information about human life,aging,disease and death are hidden in genes.Disease is the biggest killer of human health.Finding the relationship between genes and diseases will benefit human beings.In this paper,based on the previous phenotypic association data set,noncoding gene miRNA association data set and phenotypic gene association data set,the relationship between pathogenic gene and disease phenotype is mined by designing algorithm.The specific work is summarized as follows:(1)The structural characteristics of homozygosity and heterozygosity of three networks,which are composed of phenotype association data set,noncoding gene miRNA association data set and phenotype gene association data set,are distinguished.Through the identification of the structural characteristics of the three networks: the three networks are all heterozygous networks,and the nodes with large degree value tend to connect the nodes with small degree value.At the same time,combining the degree value and heterozygosity analysis of nodes in the phenotypic gene association network,we know that the same phenotype may be caused by different genes,while the same gene may lead to different phenotypes.Therefore,it is predicted that gene nodes with higher node degree may be connected with more phenotype nodes.(2)The phenotype gene double coupling network is composed of phenotype correlation data set,non coding gene miRNA correlation data set and phenotype gene correlation data set.Four meta paths are defined on the two-level coupling network of phenotype and gene.On the basis of meta path,balanced two-way random walk is carried out.The probability of walk from phenotype node to gene node is taken as the correlation value of phenotype and gene node.The correlation value of phenotype and gene existing at the same time of four meta paths is weighted and summed to get the final correlation value of phenotype and gene and sorted in descending order The first k was selected as the correlation value between phenotype and gene.Finally,the paper selects several algorithms to compare with them,and finds that the prediction effect of the algorithm proposed in the paper is better than that of other algorithms,and has some advantages.