Tissue-Specific Gene Mining Based On PET Image Phenotype

Alzheimer’s disease(AD)is a complex neurodegenerative disease.Due to the complexity of human cognition and brain function,its pathogenesis is not clear.Patients generally can not get effective treatment.At present,they can only delay the disease by intervening in the early stage of the disease course.With the increasing aging of the population in China,the incidence rate of ad is also rising,which brings great pressure to every family and even the whole society.Therefore,it is very important to find effective early prediction methods for AD.Since it was proposed,genome wide association analysis(GWAS)has become an important means to detect ad,and has successfully located more than 20 AD risk variation sites.However,it only studies the relationship between a single site and phenotype without considering the relationship between sites.Therefore,a random association analysis method based on biological network has been born,which uses statistical means,By introducing gene or protein interaction network,the results of GWAS are reordered to mine more phenotypic site information.This strategy is named NetWAS.In this paper,a NetWAS analysis method based on fully connected neural network is studied.By introducing tissue-specific functional interaction network,the genes related to petav45 phenotype of frontal lobe,parietal lobe and temporal lobe of human cerebral cortex are mined.Firstly,the whole genome association analysis of pet-av45 phenotype in the above three brain regions was carried out to obtain the list of associated genes and the association value pvalue.The four significantly associated genes found in the preliminary analysis proved the rationality of data screening and GWAS results.Secondly,the neural network is constructed by using the GWAS p-value level as the input of the neural network to predict the function of the whole gene network,and then the GWAS p-value level is extracted as the input of the neural network.The first 50 genes were selected as candidate genes,and the genes strongly associated with AD were excavated through comparative analysis of literature.This paper studies the netwas analysis method based on support vector regression(SVR)and ridge regression,and compares the consistency between the candidate genes obtained by the three strategies and the determined high-risk genes of AD,as well as the functional interaction intensity between genes.The experimental results show that the netwas analysis method based on fully connected neural network is the best.Finally,the first 50 genes in the reordered gene list were selected as candidate genes for GO term enrichment analysis and KEGG pathway analysis,and several pathways and biological processes significantly associated with the disease process of AD were obtained.In conclusion,compared with the traditional GWAS method,the netwas analysis method based on fully connected neural network studied in this paper can mine new phenotype related genes by integrating the correlation between genes,then it provides a new method for exploring the pathogenesis and development of AD.