Bidirectional Association Study Of Imaging Quantitative Phenotype And Genome-Wide SNP On Alzheimer’s Disease

Alzheimer’s Disease(AD)is an age-related neurological disease,which is clinically manifested as a gradual deterioration of cognitive function.Many studies have shown that the structure,function,and abnormality(e.g.,those related to Alzheimer’s disease)of the brain are heritable.However,which genetic variations contribute to these phenotypic changes is not completely clear.Brain image genetics provides enormous opportunities for examining the effects of genetic variations on the brain such as single nucleotide polymorphism(SNP)that affect brain structure.The voxel-wise Genome-Wide Association Studies(vGWAS)link SNP and other genetic variants with imaging phenotypes and analyze the correlation between SNP and region of interest(Region of Interest,ROI)measurements.This method is used to refine the phenotype into voxels,which greatly increases the number of candidate SNPs and genes,and can mine SNP from a subtler perspective,capturing subtle signals that are easily missed by ROI-based methods.Therefore,subdividing the hippocampus into voxel and using hippocampal voxel as a phenotype to study the explainable connection between voxel can better understand the effect of voxel set on the hippocampus and diseases.The difficulties and problems in AD research were summarized in this article and voxel were taken as the research focus.Based on voxel image data and genetic data,SNP mining model and voxel mining model were proposed,mining the abnormal brain areas and important genetic markers related to AD using different perspectives and different strategies.The research in this article focuses on bidirectional association study between voxels and SNPs,namely,the research of mining SNP based on voxel,the research of mining voxel based on SNP,the research of mining pathogenic hippocampus and abnormal genes based on the fusion features of hippocampal subregions and genes.In the data preprocessing stage,quality control was performed on the data obtained from ADNI.The participants after quality control were set as the initial group,and the selected part from the original group was used as the verification group.The image data and genetic data after quality control were combined for vGWAS to prepare data for subsequent research.The main research contents are as follows:(1)According to the vGWAS results of the initial group,in response to the large-scale data problem of the common effect of multiple markers,different strategies and multiple SNP mining model were proposed to mine subtle SNPs on voxel level and ROI level.The weight indicator was defined to balance coverage and significance using the median balance method,and then the final SNPs obtained were analyzed to find the subtle SNPs related to AD which were easily missing in traditional GWAS.(2)According to the vGWAS results of the initial group,in view of the correlation of voxels,linkage disequilibrium score regression(LDSC)was introduced to obtain the genetic correlation between voxel pairs.On this basis,a voxel-based AD genetic correlation network was discovered.The overlapping relationship between the network and the current hippocampal subregions were analyzed,and the enrichment function,pathway,and correlation with AD were also analyzed from the biological perspective.The vGWAS results of the validation group were used to reproduce the network proposed in this article,and re-verified that the network played an important role in understanding and studying AD,and the correlation between the network and AD was superior to the correlation between the current hippocampal subregions and AD.(3)According to the vGWAS data of the validation group,for the lack of single source data features’classification accuracy,a fusion feature construction method based on vGWAS results and genes was proposed.The correlation between vGWAS results and gene coding were applied to further amplify the difference between the AD group and the HC group,making it easier to detect related genes and regions.According to the fusion features obtained in(3),aiming at the accuracy of classification and accuracy of fusion features,a model for mining fusion features was proposed and constructed.The ideas of genetic evolution and cluster evolution were introduced to increase the classification accuracy of the model.Most of the hippocampal subregions and genes identified by this model overlapped with previous studies,and some were related to AD.From the biological perspective,the identified genes were analyzed,and multiple pathways related to AD were found.Based on the results of the above research,a visualization tool was designed and constructed to integrate heatmap plot,manhattan plot and brain image,realizing the association visualization of the brain image module and the statistical analysis graph.This tool provides preliminary mining and auxiliary analysis of the result data for researchers.The voxel-based research in this article can not only mine subtle SNPs that were easily missed by traditional methods,but also identify highly genetic correlation voxel set on SNP level.And pathogenic genes and abnormal hippocampal subregions were mined using the research at the same time.Studying the genetic variation and voxel-based image of the hippocampus in AD can analyze and help understanding the genetic effect and underlying affected regions of AD in the hippocampus from the biological perspective.